Merge branch 'RooCodeInc:main' into feat/fix-bedrock

2026-09-05 08:10:14 +00:00 · 2026-04-09 20:42:26 +07:00 · 2026-04-09 20:42:26 +07:00 · f89044e295
commit f89044e295
parent 3d440a7ce8 7adbfec2a4
113 changed files with 3416 additions and 769 deletions
--- a/.changeset/add-openai-gpt-5-4-mini-nano.md
+++ b/.changeset/add-openai-gpt-5-4-mini-nano.md
@ -1,6 +0,0 @@
---
-"roo-cline": patch
-"@roo-code/types": patch
---
-
-Add support for OpenAI `gpt-5.4-mini` and `gpt-5.4-nano` models.
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@ -1,5 +1,20 @@
 # Roo Code Changelog

+## 3.52.0
+
+### Minor Changes
+
+- Add Poe as an AI provider so users can access Poe models directly in Roo Code (PR #12015 by @kamilio)
+- Improve the xAI provider by migrating it to the Responses API with reusable transform utilities (#11961 by @carlesso, PR #11962 by @carlesso)
+- Fix MiniMax model listings and context window handling for more reliable configuration (#11999 by @Rexarrior, PR #12069 by @Rexarrior)
+- Add xAI Grok-4.20 models and update the default xAI model selection (#11955 by @carlesso, PR #11956 by @carlesso)
+- Add OpenAI GPT-5.4 mini and nano models to expand the available OpenAI model lineup (PR #11946 by @PeterDaveHello)
+- Chore: include the automated version bump PR from the previous release cycle for complete release accounting (PR #11892 by @app/github-actions)
+
+### Patch Changes
+
+- Add support for OpenAI `gpt-5.4-mini` and `gpt-5.4-nano` models.
+
 ## 3.51.1

 ### Patch Changes
--- a/README.md
+++ b/README.md
@ -13,10 +13,11 @@

 > Your AI-Powered Dev Team, Right in Your Editor

-## What's New in v3.51.0
+## What's New in v3.52.0

- Add support for OpenAI GPT-5.4 and GPT-5.3 Chat Latest so you can use the newest OpenAI chat models in Roo Code.
- Expose skills as slash commands with fallback execution to make reusable workflows faster to trigger.
+- Add Poe as an AI provider so you can access Poe models directly in Roo Code.
+- Improve the xAI provider with a Responses API migration, reusable transform utilities, and updated Grok-4.20 defaults.
+- Fix MiniMax model listings and context window handling for more reliable setup.

 <details>
  <summary>🌐 Available languages</summary>
@ -40,7 +41,7 @@
 - [简体中文](locales/zh-CN/README.md)
 - [繁體中文](locales/zh-TW/README.md)
 - ...
-      </details>
+  </details>

 ---

--- a/apps/web-evals/CHANGELOG.md
+++ b/apps/web-evals/CHANGELOG.md
@ -0,0 +1,3 @@
+# @roo-code/web-evals
+
+## 0.0.1
--- a/apps/web-evals/package.json
+++ b/apps/web-evals/package.json
@ -1,6 +1,6 @@
 {
 	"name": "@roo-code/web-evals",
-	"version": "0.0.0",
+	"version": "0.0.1",
 	"type": "module",
 	"scripts": {
 		"lint": "eslint src --ext=ts,tsx --max-warnings=0",
--- a/apps/web-roo-code/CHANGELOG.md
+++ b/apps/web-roo-code/CHANGELOG.md
@ -0,0 +1,3 @@
+# @roo-code/web-roo-code
+
+## 0.0.1
--- a/apps/web-roo-code/package.json
+++ b/apps/web-roo-code/package.json
@ -1,6 +1,6 @@
 {
 	"name": "@roo-code/web-roo-code",
-	"version": "0.0.0",
+	"version": "0.0.1",
 	"type": "module",
 	"scripts": {
 		"lint": "eslint src --ext=ts,tsx --max-warnings=0",
@ -18,7 +18,7 @@
 		"@radix-ui/react-navigation-menu": "^1.2.14",
 		"@radix-ui/react-slot": "^1.2.4",
 		"@roo-code/evals": "workspace:^",
-		"@roo-code/types": "^1.108.0",
+		"@roo-code/types": "workspace:^",
 		"@tanstack/react-query": "^5.90.20",
 		"@vercel/og": "^0.8.6",
 		"class-variance-authority": "^0.7.1",
--- a/locales/ca/README.md
+++ b/locales/ca/README.md
@ -13,10 +13,11 @@

 > El teu equip de desenvolupament impulsat per IA, directament al teu editor

-## Novetats de la v3.51.0
+## Novetats de la v3.52.0

- S'ha afegit compatibilitat amb OpenAI GPT-5.4 i GPT-5.3 Chat Latest perquè puguis utilitzar els models de xat més nous d'OpenAI a Roo Code.
- Les skills es poden exposar com a ordres slash amb execució de reserva perquè els fluxos de treball reutilitzables s'activin més de pressa.
+- S'ha afegit Poe com a proveïdor d'IA perquè puguis accedir directament als models de Poe a Roo Code.
+- S'ha millorat el proveïdor xAI amb una migració a la Responses API, utilitats de transformació reutilitzables i els valors per defecte actualitzats de Grok-4.20.
+- S'han corregit els llistats de models de MiniMax i la gestió de la finestra de context per oferir una configuració més fiable.

 <details>
  <summary>🌐 Idiomes disponibles</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/de/README.md
+++ b/locales/de/README.md
@ -13,10 +13,11 @@

 > Dein KI-gestütztes Dev-Team – direkt in deinem Editor

-## Neu in v3.51.0
+## Neu in v3.52.0

- Unterstützung für OpenAI GPT-5.4 und GPT-5.3 Chat Latest hinzugefügt, damit du die neuesten OpenAI-Chatmodelle in Roo Code nutzen kannst.
- Skills als Slash-Commands mit Fallback-Ausführung verfügbar gemacht, damit wiederverwendbare Workflows schneller ausgelöst werden können.
+- Poe als KI-Anbieter hinzugefügt, damit du direkt in Roo Code auf Poe-Modelle zugreifen kannst.
+- Den xAI-Anbieter mit einer Migration auf die Responses API, wiederverwendbaren Transformations-Utilities und aktualisierten Grok-4.20-Standards verbessert.
+- MiniMax-Modelllisten und die Handhabung des Kontextfensters korrigiert, damit die Einrichtung zuverlässiger funktioniert.

 <details>
  <summary>🌐 Verfügbare Sprachen</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/es/README.md
+++ b/locales/es/README.md
@ -13,10 +13,11 @@

 > Tu equipo de desarrollo con IA, directamente en tu editor

-## Novedades de la v3.51.0
+## Novedades de la v3.52.0

- Se añadió compatibilidad con OpenAI GPT-5.4 y GPT-5.3 Chat Latest para que puedas usar los modelos de chat más recientes de OpenAI en Roo Code.
- Las skills ahora pueden exponerse como comandos slash con ejecución de respaldo para activar más rápido los flujos de trabajo reutilizables.
+- Se añadió Poe como proveedor de IA para que puedas acceder directamente a los modelos de Poe en Roo Code.
+- Se mejoró el proveedor xAI con una migración a la Responses API, utilidades de transformación reutilizables y valores predeterminados actualizados de Grok-4.20.
+- Se corrigieron los listados de modelos de MiniMax y el manejo de la ventana de contexto para una configuración más fiable.

 <details>
  <summary>🌐 Idiomas disponibles</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/fr/README.md
+++ b/locales/fr/README.md
@ -13,10 +13,11 @@

 > Ton équipe de dev propulsée par l'IA, directement dans ton éditeur

-## Nouveautés de la v3.51.0
+## Nouveautés de la v3.52.0

- Ajout de la prise en charge d’OpenAI GPT-5.4 et de GPT-5.3 Chat Latest pour que tu puisses utiliser les modèles de chat OpenAI les plus récents dans Roo Code.
- Les skills peuvent désormais être exposées comme slash commands avec une exécution de secours pour déclencher plus vite des workflows réutilisables.
+- Ajout de Poe comme fournisseur d’IA pour que tu puisses accéder directement aux modèles Poe dans Roo Code.
+- Amélioration du fournisseur xAI avec une migration vers la Responses API, des utilitaires de transformation réutilisables et des valeurs par défaut Grok-4.20 mises à jour.
+- Correction des listes de modèles MiniMax et de la gestion de la fenêtre de contexte pour une configuration plus fiable.

 <details>
  <summary>🌐 Langues disponibles</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/hi/README.md
+++ b/locales/hi/README.md
@ -13,10 +13,11 @@

 > तुम्हारी AI-संचालित डेवलपमेंट टीम, सीधे तुम्हारे एडिटर में

-## v3.51.0 में नया क्या है
+## v3.52.0 में नया क्या है

- OpenAI GPT-5.4 और GPT-5.3 Chat Latest के लिए सपोर्ट जोड़ा गया है, ताकि आप Roo Code में OpenAI के सबसे नए chat models इस्तेमाल कर सकें।
- अब skills को slash commands के रूप में fallback execution के साथ expose किया जा सकता है, जिससे reusable workflows को तेज़ी से trigger किया जा सके।
+- Poe को AI provider के रूप में जोड़ा गया है, ताकि आप Roo Code में सीधे Poe models तक पहुँच सकें।
+- xAI provider को Responses API migration, reusable transform utilities, और updated Grok-4.20 defaults के साथ बेहतर बनाया गया है।
+- ज़्यादा भरोसेमंद setup के लिए MiniMax model listings और context window handling को ठीक किया गया है।

 <details>
  <summary>🌐 उपलब्ध भाषाएँ</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/id/README.md
+++ b/locales/id/README.md
@ -13,10 +13,11 @@

 > Tim dev bertenaga AI-mu, langsung di editor kamu

-## Yang Baru di v3.51.0
+## Yang Baru di v3.52.0

- Menambahkan dukungan untuk OpenAI GPT-5.4 dan GPT-5.3 Chat Latest agar kamu bisa memakai model chat OpenAI terbaru di Roo Code.
- Skill kini bisa diekspos sebagai slash command dengan eksekusi fallback supaya workflow yang bisa dipakai ulang lebih cepat dijalankan.
+- Menambahkan Poe sebagai provider AI agar kamu bisa langsung mengakses model Poe di Roo Code.
+- Meningkatkan provider xAI dengan migrasi ke Responses API, utilitas transformasi yang bisa dipakai ulang, dan default Grok-4.20 yang diperbarui.
+- Memperbaiki daftar model MiniMax dan penanganan context window agar penyiapannya lebih andal.

 <details>
  <summary>🌐 Bahasa yang tersedia</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/it/README.md
+++ b/locales/it/README.md
@ -13,10 +13,11 @@

 > Il tuo team di sviluppo con IA, direttamente nel tuo editor

-## Novità in v3.51.0
+## Novità in v3.52.0

- Aggiunto il supporto per OpenAI GPT-5.4 e GPT-5.3 Chat Latest così puoi usare i modelli di chat OpenAI più recenti in Roo Code.
- Le skill ora possono essere esposte come slash command con esecuzione di fallback per attivare più rapidamente i workflow riutilizzabili.
+- Aggiunto Poe come provider AI così puoi accedere direttamente ai modelli Poe in Roo Code.
+- Migliorato il provider xAI con una migrazione alla Responses API, utility di trasformazione riutilizzabili e valori predefiniti aggiornati per Grok-4.20.
+- Corretti gli elenchi dei modelli MiniMax e la gestione della finestra di contesto per una configurazione più affidabile.

 <details>
  <summary>🌐 Lingue disponibili</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/ja/README.md
+++ b/locales/ja/README.md
@ -13,10 +13,11 @@

 > あなたのエディタの中に、AIで強化された開発チームを

-## v3.51.0 の新機能
+## v3.52.0 の新機能

- OpenAI GPT-5.4 と GPT-5.3 Chat Latest のサポートを追加し、Roo Code で OpenAI の最新チャットモデルを使えるようにしました。
- スキルをスラッシュコマンドとして公開し、フォールバック実行にも対応したことで、再利用可能なワークフローをより素早く呼び出せます。
+- Poe を AI プロバイダーとして追加し、Roo Code から Poe のモデルに直接アクセスできるようにしました。
+- xAI プロバイダーを Responses API への移行、再利用可能な変換ユーティリティ、更新された Grok-4.20 のデフォルト設定で改善しました。
+- より信頼性の高いセットアップのために、MiniMax のモデル一覧とコンテキストウィンドウの扱いを修正しました。

 <details>
  <summary>🌐 利用可能な言語</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/ko/README.md
+++ b/locales/ko/README.md
@ -13,10 +13,11 @@

 > AI로 강화된 너의 개발 팀, 네 에디터 안에

-## v3.51.0의 새로운 기능
+## v3.52.0의 새로운 기능

- OpenAI GPT-5.4와 GPT-5.3 Chat Latest 지원을 추가해 Roo Code에서 최신 OpenAI 채팅 모델을 사용할 수 있어요.
- 이제 스킬을 슬래시 명령어로 노출하고 fallback 실행도 지원해 재사용 가능한 워크플로를 더 빠르게 트리거할 수 있어요.
+- Poe를 AI 제공업체로 추가해 Roo Code에서 Poe 모델에 직접 접근할 수 있게 했어요.
+- Responses API 마이그레이션, 재사용 가능한 변환 유틸리티, 업데이트된 Grok-4.20 기본값으로 xAI 제공업체를 개선했어요.
+- 더 안정적인 설정을 위해 MiniMax 모델 목록과 컨텍스트 윈도우 처리를 수정했어요.

 <details>
  <summary>🌐 사용 가능한 언어</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/nl/README.md
+++ b/locales/nl/README.md
@ -13,10 +13,11 @@

 > Jouw AI-aangedreven dev-team, rechtstreeks in je editor

-## Nieuw in v3.51.0
+## Nieuw in v3.52.0

- Ondersteuning toegevoegd voor OpenAI GPT-5.4 en GPT-5.3 Chat Latest, zodat je de nieuwste OpenAI-chatmodellen in Roo Code kunt gebruiken.
- Skills beschikbaar gemaakt als slash-commands met fallback-uitvoering, zodat herbruikbare workflows sneller te starten zijn.
+- Poe toegevoegd als AI-provider, zodat je in Roo Code direct toegang hebt tot Poe-modellen.
+- De xAI-provider verbeterd met een migratie naar de Responses API, herbruikbare transformatiehulpmiddelen en bijgewerkte Grok-4.20-standaardinstellingen.
+- MiniMax-modellijsten en de afhandeling van het contextvenster gerepareerd voor een betrouwbaardere configuratie.

 <details>
  <summary>🌐 Beschikbare talen</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/pl/README.md
+++ b/locales/pl/README.md
@ -13,10 +13,11 @@

 > Twój zespół deweloperski zasilany AI — prosto w edytorze

-## Nowości w v3.51.0
+## Nowości w v3.52.0

- Dodano obsługę OpenAI GPT-5.4 i GPT-5.3 Chat Latest, żebyś mógł używać najnowszych modeli czatu OpenAI w Roo Code.
- Skills można teraz udostępniać jako slash commandy z wykonaniem awaryjnym, żeby szybciej uruchamiać wielokrotnego użytku workflowy.
+- Dodano Poe jako dostawcę AI, żebyś mógł bezpośrednio uzyskać dostęp do modeli Poe w Roo Code.
+- Ulepszono dostawcę xAI dzięki migracji do Responses API, narzędziom transformacji wielokrotnego użytku i zaktualizowanym domyślnym ustawieniom Grok-4.20.
+- Naprawiono listy modeli MiniMax i obsługę okna kontekstu, żeby konfiguracja była bardziej niezawodna.

 <details>
  <summary>🌐 Dostępne języki</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/pt-BR/README.md
+++ b/locales/pt-BR/README.md
@ -13,10 +13,11 @@

 > Seu time de desenvolvimento com IA, direto no seu editor

-## Novidades na v3.51.0
+## Novidades na v3.52.0

- Adicionamos suporte ao OpenAI GPT-5.4 e ao GPT-5.3 Chat Latest para você usar os modelos de chat mais novos da OpenAI no Roo Code.
- Agora as skills podem ser expostas como comandos slash com execução de fallback para acionar workflows reutilizáveis mais rápido.
+- Adicionamos o Poe como provedor de IA para você acessar os modelos do Poe diretamente no Roo Code.
+- Melhoramos o provedor xAI com uma migração para a Responses API, utilitários de transformação reutilizáveis e padrões atualizados do Grok-4.20.
+- Corrigimos as listagens de modelos do MiniMax e o tratamento da janela de contexto para uma configuração mais confiável.

 <details>
  <summary>🌐 Idiomas disponíveis</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/ru/README.md
+++ b/locales/ru/README.md
@ -13,10 +13,11 @@

 > Твоя команда разработчиков на ИИ — прямо в редакторе

-## Что нового в v3.51.0
+## Что нового в v3.52.0

- Добавлена поддержка OpenAI GPT-5.4 и GPT-5.3 Chat Latest, чтобы ты мог использовать новейшие чат-модели OpenAI в Roo Code.
- Skills теперь можно открывать как slash-команды с резервным выполнением, чтобы быстрее запускать переиспользуемые рабочие процессы.
+- Добавлен Poe как AI-провайдер, чтобы ты мог напрямую получать доступ к моделям Poe в Roo Code.
+- Улучшен провайдер xAI: выполнена миграция на Responses API, добавлены переиспользуемые утилиты трансформации и обновлены настройки Grok-4.20 по умолчанию.
+- Исправлены списки моделей MiniMax и обработка окна контекста для более надёжной настройки.

 <details>
  <summary>🌐 Доступные языки</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/tr/README.md
+++ b/locales/tr/README.md
@ -13,10 +13,11 @@

 > AI destekli dev ekibin, doğrudan editörünün içinde

-## v3.51.0'daki Yenilikler
+## v3.52.0'daki Yenilikler

- Roo Code içinde en yeni OpenAI sohbet modellerini kullanabilmen için OpenAI GPT-5.4 ve GPT-5.3 Chat Latest desteği eklendi.
- Skills artık fallback yürütmeyle slash komutları olarak sunulabiliyor, böylece yeniden kullanılabilir iş akışları daha hızlı tetiklenebiliyor.
+- Poe, Roo Code içinde Poe modellerine doğrudan erişebilmen için bir AI sağlayıcısı olarak eklendi.
+- xAI sağlayıcısı, Responses API geçişi, yeniden kullanılabilir dönüştürme yardımcıları ve güncellenmiş Grok-4.20 varsayılanlarıyla iyileştirildi.
+- Daha güvenilir bir kurulum için MiniMax model listeleri ve context window işleme düzeltildi.

 <details>
  <summary>🌐 Mevcut diller</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/vi/README.md
+++ b/locales/vi/README.md
@ -13,10 +13,11 @@

 > Đội ngũ dev dùng AI của bạn, ngay trong trình chỉnh sửa

-## Điểm mới trong v3.51.0
+## Điểm mới trong v3.52.0

- Đã thêm hỗ trợ cho OpenAI GPT-5.4 và GPT-5.3 Chat Latest để bạn có thể dùng các mô hình chat OpenAI mới nhất trong Roo Code.
- Skills giờ có thể được cung cấp dưới dạng lệnh slash kèm thực thi dự phòng để kích hoạt các quy trình tái sử dụng nhanh hơn.
+- Đã thêm Poe làm nhà cung cấp AI để bạn có thể truy cập trực tiếp các model Poe trong Roo Code.
+- Đã cải thiện nhà cung cấp xAI với quá trình chuyển sang Responses API, các tiện ích chuyển đổi có thể tái sử dụng và các giá trị mặc định Grok-4.20 đã được cập nhật.
+- Đã sửa danh sách model MiniMax và cách xử lý context window để thiết lập đáng tin cậy hơn.

 <details>
  <summary>🌐 Các ngôn ngữ có sẵn</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/zh-CN/README.md
+++ b/locales/zh-CN/README.md
@ -13,10 +13,11 @@

 > 你的 AI 驱动开发团队，就在你的编辑器里

-## v3.51.0 新增内容
+## v3.52.0 新增内容

- 已添加对 OpenAI GPT-5.4 和 GPT-5.3 Chat Latest 的支持，让你可以在 Roo Code 中使用 OpenAI 最新的聊天模型。
- 现在可以将 skills 公开为 slash 命令，并支持 fallback execution，让可复用工作流触发得更快。
+- 已将 Poe 添加为 AI 提供商，让你可以在 Roo Code 中直接访问 Poe 模型。
+- 已改进 xAI 提供商，完成了向 Responses API 的迁移，加入了可复用的转换工具，并更新了 Grok-4.20 默认设置。
+- 已修复 MiniMax 模型列表和上下文窗口处理方式，让配置更加可靠。

 <details>
  <summary>🌐 可用语言</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-  </details>
+    </details>

 ---

--- a/locales/zh-TW/README.md
+++ b/locales/zh-TW/README.md
@ -13,10 +13,11 @@

 > 您的 AI 驅動開發團隊，就在您的編輯器中

-## v3.51.0 新功能
+## v3.52.0 新功能

- 已新增對 OpenAI GPT-5.4 和 GPT-5.3 Chat Latest 的支援，讓你可以在 Roo Code 中使用 OpenAI 最新的聊天模型。
- 現在可以將 skills 公開為 slash 指令，並支援 fallback execution，讓可重用工作流程觸發得更快。
+- 已將 Poe 新增為 AI 供應商，讓你可以在 Roo Code 中直接存取 Poe 模型。
+- 已改善 xAI 供應商，完成 Responses API 遷移、加入可重複使用的轉換工具，並更新了 Grok-4.20 預設值。
+- 已修正 MiniMax 模型清單和上下文視窗處理方式，讓設定更加可靠。

 <details>
  <summary>🌐 支援語言</summary>
@ -40,7 +41,7 @@
 - [简体中文](../zh-CN/README.md)
 - [繁體中文](../zh-TW/README.md)
 - ...
-      </details>
+  </details>

 ---

--- a/packages/cloud/CHANGELOG.md
+++ b/packages/cloud/CHANGELOG.md
@ -0,0 +1,3 @@
+# @roo-code/cloud
+
+## 0.0.1
--- a/packages/cloud/package.json
+++ b/packages/cloud/package.json
@ -1,7 +1,7 @@
 {
 	"name": "@roo-code/cloud",
 	"description": "Roo Code Cloud services.",
-	"version": "0.0.0",
+	"version": "0.0.1",
 	"type": "module",
 	"exports": "./src/index.ts",
 	"scripts": {
--- a/packages/core/CHANGELOG.md
+++ b/packages/core/CHANGELOG.md
@ -0,0 +1,3 @@
+# @roo-code/core
+
+## 0.0.1
--- a/packages/core/package.json
+++ b/packages/core/package.json
@ -1,7 +1,7 @@
 {
 	"name": "@roo-code/core",
 	"description": "Platform agnostic core functionality for Roo Code.",
-	"version": "0.0.0",
+	"version": "0.0.1",
 	"type": "module",
 	"exports": {
 		".": "./src/index.ts",
--- a/packages/evals/CHANGELOG.md
+++ b/packages/evals/CHANGELOG.md
@ -0,0 +1,3 @@
+# @roo-code/evals
+
+## 0.0.1
--- a/packages/evals/package.json
+++ b/packages/evals/package.json
@ -1,7 +1,7 @@
 {
 	"name": "@roo-code/evals",
 	"description": "Roo Code evals.",
-	"version": "0.0.0",
+	"version": "0.0.1",
 	"type": "module",
 	"exports": "./src/index.ts",
 	"scripts": {
--- a/packages/ipc/CHANGELOG.md
+++ b/packages/ipc/CHANGELOG.md
@ -0,0 +1,3 @@
+# @roo-code/ipc
+
+## 0.0.1
--- a/packages/ipc/package.json
+++ b/packages/ipc/package.json
@ -1,7 +1,7 @@
 {
 	"name": "@roo-code/ipc",
 	"description": "IPC server and client for remote Roo Code access.",
-	"version": "0.0.0",
+	"version": "0.0.1",
 	"type": "module",
 	"exports": "./src/index.ts",
 	"scripts": {
--- a/packages/telemetry/CHANGELOG.md
+++ b/packages/telemetry/CHANGELOG.md
@ -0,0 +1,3 @@
+# @roo-code/telemetry
+
+## 0.0.1
--- a/packages/telemetry/package.json
+++ b/packages/telemetry/package.json
@ -1,7 +1,7 @@
 {
 	"name": "@roo-code/telemetry",
 	"description": "Roo Code telemetry service and clients.",
-	"version": "0.0.0",
+	"version": "0.0.1",
 	"type": "module",
 	"exports": "./src/index.ts",
 	"scripts": {
--- a/packages/types/CHANGELOG.md
+++ b/packages/types/CHANGELOG.md
@ -0,0 +1,7 @@
+# @roo-code/types
+
+## 0.0.1
+
+### Patch Changes
+
+- Add support for OpenAI `gpt-5.4-mini` and `gpt-5.4-nano` models.
--- a/packages/types/package.json
+++ b/packages/types/package.json
@ -1,6 +1,6 @@
 {
 	"name": "@roo-code/types",
-	"version": "0.0.0",
+	"version": "0.0.1",
 	"type": "module",
 	"main": "./dist/index.cjs",
 	"exports": {
@ -20,9 +20,11 @@
 		"build": "tsup",
 		"build:watch": "tsup --watch --outDir npm/dist --onSuccess 'echo ✅ Types rebuilt to npm/dist'",
 		"npm:publish": "node scripts/publish-npm.cjs",
-		"clean": "rimraf dist .turbo"
+		"clean": "rimraf dist .turbo",
+		"generate:schema": "tsx scripts/generate-roomodes-schema.ts"
 	},
 	"dependencies": {
+		"ai-sdk-provider-poe": "2.0.18",
 		"zod": "3.25.76"
 	},
 	"devDependencies": {
@ -31,6 +33,8 @@
 		"@types/node": "^24.1.0",
 		"globals": "^16.3.0",
 		"tsup": "^8.4.0",
-		"vitest": "^3.2.3"
+		"ajv": "^8.18.0",
+		"vitest": "^3.2.3",
+		"zod-to-json-schema": "^3.25.1"
 	}
 }
--- a/packages/types/scripts/generate-roomodes-schema.ts
+++ b/packages/types/scripts/generate-roomodes-schema.ts
@ -0,0 +1,25 @@
+/**
+ * Generates the JSON Schema for .roomodes configuration files from the Zod
+ * schemas defined in packages/types/src/mode.ts.
+ *
+ * This ensures the schema stays in sync with the TypeScript types. Run via:
+ *   pnpm --filter @roo-code/types generate:schema
+ *
+ * The output is written to schemas/roomodes.json at the repository root.
+ */
+
+import * as fs from "fs"
+import * as path from "path"
+import { fileURLToPath } from "url"
+
+import { generateRoomodesJsonSchema } from "../src/roomodes-schema.js"
+
+const jsonSchema = generateRoomodesJsonSchema()
+
+const __dirname = path.dirname(fileURLToPath(import.meta.url))
+const repoRoot = path.resolve(__dirname, "../../..")
+const outPath = path.join(repoRoot, "schemas", "roomodes.json")
+fs.mkdirSync(path.dirname(outPath), { recursive: true })
+fs.writeFileSync(outPath, JSON.stringify(jsonSchema, null, "\t") + "\n", "utf-8")
+
+console.log(`Generated ${path.relative(repoRoot, outPath)}`)
--- a/packages/types/src/tests/roomodes-schema-sync.spec.ts
+++ b/packages/types/src/tests/roomodes-schema-sync.spec.ts
@ -0,0 +1,26 @@
+import { describe, it, expect } from "vitest"
+import * as fs from "fs"
+import * as path from "path"
+import { fileURLToPath } from "url"
+
+import { generateRoomodesJsonSchema } from "../roomodes-schema.js"
+
+/**
+ * This test verifies that the checked-in schemas/roomodes.json matches what
+ * would be generated from the current Zod schemas. If this test fails, run:
+ *
+ *   pnpm --filter @roo-code/types generate:schema
+ *
+ * to regenerate the schema file.
+ */
+describe("roomodes schema sync", () => {
+	it("should match the dynamically generated schema from Zod types", () => {
+		const __dirname = path.dirname(fileURLToPath(import.meta.url))
+		const schemaPath = path.resolve(__dirname, "../../../../schemas/roomodes.json")
+		const checkedIn = JSON.parse(fs.readFileSync(schemaPath, "utf-8"))
+
+		const generated = generateRoomodesJsonSchema()
+
+		expect(checkedIn).toEqual(generated)
+	})
+})
--- a/packages/types/src/tests/roomodes-schema.spec.ts
+++ b/packages/types/src/tests/roomodes-schema.spec.ts
@ -0,0 +1,438 @@
+/**
+ * Validates the generated schemas/roomodes.json against sample configurations
+ * using AJV. The schema itself is dynamically generated from the Zod types in
+ * packages/types/src/mode.ts -- see packages/types/scripts/generate-roomodes-schema.ts.
+ *
+ * A separate drift-detection test (roomodes-schema-sync.spec.ts) ensures the
+ * checked-in schema stays in sync with the Zod source of truth.
+ */
+import { describe, it, expect, beforeAll } from "vitest"
+import Ajv, { type ValidateFunction } from "ajv"
+import * as fs from "fs"
+import * as path from "path"
+import { fileURLToPath } from "url"
+
+describe("roomodes JSON schema", () => {
+	let schema: Record<string, unknown>
+	let validate: ValidateFunction
+
+	beforeAll(() => {
+		const __dirname = path.dirname(fileURLToPath(import.meta.url))
+		const schemaPath = path.resolve(__dirname, "../../../../schemas/roomodes.json")
+		schema = JSON.parse(fs.readFileSync(schemaPath, "utf-8"))
+		const ajv = new Ajv.default({ strict: false })
+		validate = ajv.compile(schema)
+	})
+
+	it("should be a valid JSON Schema", () => {
+		expect(validate).toBeDefined()
+	})
+
+	it("should accept a minimal valid .roomodes config", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "my-mode",
+					name: "My Mode",
+					roleDefinition: "You are a helpful assistant.",
+					groups: ["read"],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should accept a mode with all optional properties", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "full-mode",
+					name: "Full Mode",
+					roleDefinition: "A complete mode definition.",
+					whenToUse: "Use when you need everything.",
+					description: "A mode with all properties.",
+					customInstructions: "Follow these additional rules.",
+					groups: ["read", "edit", "command", "mcp"],
+					source: "project",
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should accept the built-in architect mode with tuple-style edit group", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "architect",
+					name: "Architect",
+					roleDefinition: "You are an experienced technical leader.",
+					whenToUse: "Use this mode when you need to plan.",
+					description: "Plan and design before implementation",
+					groups: ["read", ["edit", { fileRegex: "\\.md$", description: "Markdown files only" }], "mcp"],
+					source: "project",
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should accept a tuple group entry with only fileRegex", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "restricted",
+					name: "Restricted",
+					roleDefinition: "Limited editor.",
+					groups: [["edit", { fileRegex: "\\.ts$" }]],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should accept a tuple group entry with empty options", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "empty-opts",
+					name: "Empty Options",
+					roleDefinition: "Mode with empty group options.",
+					groups: [["edit", {}]],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should accept the modes tool group", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "orchestrator",
+					name: "Orchestrator",
+					roleDefinition: "You orchestrate other modes.",
+					groups: ["read", "modes"],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should reject a config missing customModes", () => {
+		const config = {}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should reject a mode missing required slug", () => {
+		const config = {
+			customModes: [
+				{
+					name: "No Slug",
+					roleDefinition: "Missing slug.",
+					groups: ["read"],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should reject a mode missing required groups", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "no-groups",
+					name: "No Groups",
+					roleDefinition: "Missing groups.",
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should reject a slug with invalid characters", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "invalid slug!",
+					name: "Bad Slug",
+					roleDefinition: "Invalid slug characters.",
+					groups: ["read"],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should reject an invalid tool group name", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "bad-group",
+					name: "Bad Group",
+					roleDefinition: "Invalid group name.",
+					groups: ["nonexistent"],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should reject additional properties on CustomMode", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "extra-props",
+					name: "Extra Props",
+					roleDefinition: "Has extra properties.",
+					groups: ["read"],
+					unknownField: true,
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should reject additional properties on GroupOptions", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "bad-opts",
+					name: "Bad Options",
+					roleDefinition: "Extra options properties.",
+					groups: [["edit", { fileRegex: "\\.md$", unknownOpt: true }]],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should reject a tuple with more than 2 elements", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "big-tuple",
+					name: "Big Tuple",
+					roleDefinition: "Too many tuple elements.",
+					groups: [["edit", { fileRegex: "\\.md$" }, "extra"]],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should reject an invalid source value", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "bad-source",
+					name: "Bad Source",
+					roleDefinition: "Invalid source.",
+					groups: ["read"],
+					source: "unknown",
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should accept an empty customModes array", () => {
+		const config = {
+			customModes: [],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should accept the browser tool group (deprecated but valid)", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "browser-mode",
+					name: "Browser Mode",
+					roleDefinition: "A mode that uses the browser tool group.",
+					groups: ["read", "browser", "command"],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should accept a browser tuple group entry", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "browser-tuple",
+					name: "Browser Tuple",
+					roleDefinition: "A mode with browser tuple.",
+					groups: [["browser", { fileRegex: "\\.html$", description: "HTML files only" }]],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should accept a mode with rulesFiles", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "rules-mode",
+					name: "Rules Mode",
+					roleDefinition: "A mode with rules files.",
+					groups: ["read"],
+					rulesFiles: [
+						{
+							relativePath: "rule1.md",
+							content: "# Rule 1\nFollow this rule.",
+						},
+						{
+							relativePath: "subfolder/rule2.md",
+							content: "# Rule 2\nFollow this other rule.",
+						},
+					],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should accept a mode with empty rulesFiles array", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "empty-rules",
+					name: "Empty Rules",
+					roleDefinition: "A mode with empty rules files.",
+					groups: ["read"],
+					rulesFiles: [],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should accept rulesFiles entries with only relativePath (content is optional)", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "path-only-rules",
+					name: "Path Only Rules",
+					roleDefinition: "A mode with rules files that only have relativePath.",
+					groups: ["read"],
+					rulesFiles: [{ relativePath: "rule1.md" }],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+
+	it("should reject rulesFiles entries missing required relativePath", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "bad-rules",
+					name: "Bad Rules",
+					roleDefinition: "A mode with invalid rules files.",
+					groups: ["read"],
+					rulesFiles: [{ content: "some content" }],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should reject rulesFiles entries with extra properties", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "extra-rules",
+					name: "Extra Rules",
+					roleDefinition: "A mode with extra rule properties.",
+					groups: ["read"],
+					rulesFiles: [{ relativePath: "rule1.md", content: "content", extra: true }],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(valid).toBe(false)
+	})
+
+	it("should accept multiple modes", () => {
+		const config = {
+			customModes: [
+				{
+					slug: "mode-a",
+					name: "Mode A",
+					roleDefinition: "First mode.",
+					groups: ["read"],
+				},
+				{
+					slug: "mode-b",
+					name: "Mode B",
+					roleDefinition: "Second mode.",
+					groups: ["read", "edit"],
+				},
+			],
+		}
+
+		const valid = validate(config)
+		expect(validate.errors).toBeNull()
+		expect(valid).toBe(true)
+	})
+})
--- a/packages/types/src/provider-settings.ts
+++ b/packages/types/src/provider-settings.ts
@ -34,7 +34,15 @@ export const DEFAULT_CONSECUTIVE_MISTAKE_LIMIT = 3
 * Dynamic provider requires external API calls in order to get the model list.
 */

-export const dynamicProviders = ["openrouter", "vercel-ai-gateway", "litellm", "requesty", "roo", "unbound"] as const
+export const dynamicProviders = [
+	"openrouter",
+	"vercel-ai-gateway",
+	"litellm",
+	"requesty",
+	"roo",
+	"unbound",
+	"poe",
+] as const

 export type DynamicProvider = (typeof dynamicProviders)[number]

@ -306,6 +314,11 @@ const deepSeekSchema = apiModelIdProviderModelSchema.extend({
 	deepSeekApiKey: z.string().optional(),
 })

+const poeSchema = apiModelIdProviderModelSchema.extend({
+	poeApiKey: z.string().optional(),
+	poeBaseUrl: z.string().optional(),
+})
+
 const moonshotSchema = apiModelIdProviderModelSchema.extend({
 	moonshotBaseUrl: z
 		.union([z.literal("https://api.moonshot.ai/v1"), z.literal("https://api.moonshot.cn/v1")])
@ -400,6 +413,7 @@ export const providerSettingsSchemaDiscriminated = z.discriminatedUnion("apiProv
 	openAiNativeSchema.merge(z.object({ apiProvider: z.literal("openai-native") })),
 	mistralSchema.merge(z.object({ apiProvider: z.literal("mistral") })),
 	deepSeekSchema.merge(z.object({ apiProvider: z.literal("deepseek") })),
+	poeSchema.merge(z.object({ apiProvider: z.literal("poe") })),
 	moonshotSchema.merge(z.object({ apiProvider: z.literal("moonshot") })),
 	minimaxSchema.merge(z.object({ apiProvider: z.literal("minimax") })),
 	requestySchema.merge(z.object({ apiProvider: z.literal("requesty") })),
@ -433,6 +447,7 @@ export const providerSettingsSchema = z.object({
 	...openAiNativeSchema.shape,
 	...mistralSchema.shape,
 	...deepSeekSchema.shape,
+	...poeSchema.shape,
 	...moonshotSchema.shape,
 	...minimaxSchema.shape,
 	...requestySchema.shape,
@ -510,6 +525,7 @@ export const modelIdKeysByProvider: Record<TypicalProvider, ModelIdKey> = {
 	moonshot: "apiModelId",
 	minimax: "apiModelId",
 	deepseek: "apiModelId",
+	poe: "apiModelId",
 	"qwen-code": "apiModelId",
 	requesty: "requestyModelId",
 	unbound: "unboundModelId",
@ -632,6 +648,7 @@ export const MODELS_BY_PROVIDER: Record<
 	baseten: { id: "baseten", label: "Baseten", models: Object.keys(basetenModels) },

 	// Dynamic providers; models pulled from remote APIs.
+	poe: { id: "poe", label: "Poe", models: [] },
 	litellm: { id: "litellm", label: "LiteLLM", models: [] },
 	openrouter: { id: "openrouter", label: "OpenRouter", models: [] },
 	requesty: { id: "requesty", label: "Requesty", models: [] },
--- a/packages/types/src/providers/index.ts
+++ b/packages/types/src/providers/index.ts
@ -13,6 +13,7 @@ export * from "./openai.js"
 export * from "./openai-codex.js"
 export * from "./openai-codex-rate-limits.js"
 export * from "./openrouter.js"
+export * from "./poe.js"
 export * from "./qwen-code.js"
 export * from "./requesty.js"
 export * from "./roo.js"
@ -36,6 +37,7 @@ import { mistralDefaultModelId } from "./mistral.js"
 import { moonshotDefaultModelId } from "./moonshot.js"
 import { openAiCodexDefaultModelId } from "./openai-codex.js"
 import { openRouterDefaultModelId } from "./openrouter.js"
+import { poeDefaultModelId } from "./poe.js"
 import { qwenCodeDefaultModelId } from "./qwen-code.js"
 import { requestyDefaultModelId } from "./requesty.js"
 import { rooDefaultModelId } from "./roo.js"
@ -107,6 +109,8 @@ export function getProviderDefaultModelId(
 			return rooDefaultModelId
 		case "qwen-code":
 			return qwenCodeDefaultModelId
+		case "poe":
+			return poeDefaultModelId
 		case "unbound":
 			return unboundDefaultModelId
 		case "vercel-ai-gateway":
--- a/packages/types/src/providers/minimax.ts
+++ b/packages/types/src/providers/minimax.ts
@ -1,11 +1,11 @@
 import type { ModelInfo } from "../model.js"

 // Minimax
-// https://platform.minimax.io/docs/guides/pricing
-// https://platform.minimax.io/docs/api-reference/text-openai-api
-// https://platform.minimax.io/docs/api-reference/text-anthropic-api
+// https://platform.minimax.io/docs/guides/models-intro
+// https://platform.minimax.io/docs/guides/pricing-paygo
+// https://platform.minimax.io/docs/guides/pricing-tokenplan
 export type MinimaxModelId = keyof typeof minimaxModels
-export const minimaxDefaultModelId: MinimaxModelId = "MiniMax-M2.5"
+export const minimaxDefaultModelId: MinimaxModelId = "MiniMax-M2.7"

 export const minimaxModels = {
 	"MiniMax-M2.5": {
@ -21,11 +21,56 @@ export const minimaxModels = {
 		cacheWritesPrice: 0.375,
 		cacheReadsPrice: 0.03,
 		description:
-			"MiniMax M2.5, the latest MiniMax model with enhanced coding and agentic capabilities, building on the strengths of the M2 series.",
+			"MiniMax M2.5, the latest MiniMax model with enhanced coding and agentic capabilities, building on the strengths of the M2 series. See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Note: When using TokenPlan, usage is billed per request, not per token.",
+	},
+	"MiniMax-M2.5-highspeed": {
+		maxTokens: 16_384,
+		contextWindow: 204_800,
+		supportsImages: false,
+		supportsPromptCache: true,
+		includedTools: ["search_and_replace"],
+		excludedTools: ["apply_diff"],
+		preserveReasoning: true,
+		inputPrice: 0.6,
+		outputPrice: 2.4,
+		cacheWritesPrice: 0.375,
+		cacheReadsPrice: 0.03,
+		description:
+			"MiniMax M2.5 highspeed: same performance as M2.5 but with faster response (approximately 100 tps vs 60 tps). See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Requires TokenPlan High-Speed subscription for use with TokenPlan keys. Note: When using TokenPlan, usage is billed per request, not per token.",
+	},
+	"MiniMax-M2.7": {
+		maxTokens: 16_384,
+		contextWindow: 204_800,
+		supportsImages: false,
+		supportsPromptCache: true,
+		includedTools: ["search_and_replace"],
+		excludedTools: ["apply_diff"],
+		preserveReasoning: true,
+		inputPrice: 0.3,
+		outputPrice: 1.2,
+		cacheWritesPrice: 0.375,
+		cacheReadsPrice: 0.06,
+		description:
+			"MiniMax M2.7, the latest MiniMax model with recursive self-improvement capabilities. See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Note: When using TokenPlan, usage is billed per request, not per token.",
+	},
+	"MiniMax-M2.7-highspeed": {
+		maxTokens: 16_384,
+		contextWindow: 204_800,
+		supportsImages: false,
+		supportsPromptCache: true,
+		includedTools: ["search_and_replace"],
+		excludedTools: ["apply_diff"],
+		preserveReasoning: true,
+		inputPrice: 0.6,
+		outputPrice: 2.4,
+		cacheWritesPrice: 0.375,
+		cacheReadsPrice: 0.06,
+		description:
+			"MiniMax M2.7 highspeed: same performance as M2.7 but with faster response (approximately 100 tps vs 60 tps). See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Requires TokenPlan High-Speed subscription for use with TokenPlan keys. Note: When using TokenPlan, usage is billed per request, not per token.",
 	},
 	"MiniMax-M2": {
 		maxTokens: 16_384,
-		contextWindow: 192_000,
+		contextWindow: 204_800,
 		supportsImages: false,
 		supportsPromptCache: true,
 		includedTools: ["search_and_replace"],
@ -36,11 +81,11 @@ export const minimaxModels = {
 		cacheWritesPrice: 0.375,
 		cacheReadsPrice: 0.03,
 		description:
-			"MiniMax M2, a model born for Agents and code, featuring Top-tier Coding Capabilities, Powerful Agentic Performance, and Ultimate Cost-Effectiveness & Speed.",
+			"MiniMax M2, a model born for Agents and code, featuring Top-tier Coding Capabilities, Powerful Agentic Performance, and Ultimate Cost-Effectiveness & Speed. See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Note: When using TokenPlan, usage is billed per request, not per token.",
 	},
 	"MiniMax-M2-Stable": {
 		maxTokens: 16_384,
-		contextWindow: 192_000,
+		contextWindow: 204_800,
 		supportsImages: false,
 		supportsPromptCache: true,
 		includedTools: ["search_and_replace"],
@ -51,11 +96,11 @@ export const minimaxModels = {
 		cacheWritesPrice: 0.375,
 		cacheReadsPrice: 0.03,
 		description:
-			"MiniMax M2 Stable (High Concurrency, Commercial Use), a model born for Agents and code, featuring Top-tier Coding Capabilities, Powerful Agentic Performance, and Ultimate Cost-Effectiveness & Speed.",
+			"MiniMax M2 Stable (High Concurrency, Commercial Use), a model born for Agents and code, featuring Top-tier Coding Capabilities, Powerful Agentic Performance, and Ultimate Cost-Effectiveness & Speed. See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Note: When using TokenPlan, usage is billed per request, not per token.",
 	},
 	"MiniMax-M2.1": {
 		maxTokens: 16_384,
-		contextWindow: 192_000,
+		contextWindow: 204_800,
 		supportsImages: false,
 		supportsPromptCache: true,
 		includedTools: ["search_and_replace"],
@ -66,7 +111,22 @@ export const minimaxModels = {
 		cacheWritesPrice: 0.375,
 		cacheReadsPrice: 0.03,
 		description:
-			"MiniMax M2.1 builds on M2 with improved overall performance for agentic coding tasks and significantly faster response times.",
+			"MiniMax M2.1 builds on M2 with improved overall performance for agentic coding tasks and significantly faster response times. See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Note: When using TokenPlan, usage is billed per request, not per token.",
+	},
+	"MiniMax-M2.1-highspeed": {
+		maxTokens: 16_384,
+		contextWindow: 204_800,
+		supportsImages: false,
+		supportsPromptCache: true,
+		includedTools: ["search_and_replace"],
+		excludedTools: ["apply_diff"],
+		preserveReasoning: true,
+		inputPrice: 0.6,
+		outputPrice: 2.4,
+		cacheWritesPrice: 0.375,
+		cacheReadsPrice: 0.03,
+		description:
+			"MiniMax M2.1 highspeed: same performance as M2.1 but with faster response (approximately 100 tps vs 60 tps). See pricing at https://platform.minimax.io/docs/guides/pricing-paygo. Requires TokenPlan High-Speed subscription for use with TokenPlan keys. Note: When using TokenPlan, usage is billed per request, not per token.",
 	},
 } as const satisfies Record<string, ModelInfo>

--- a/packages/types/src/providers/poe.ts
+++ b/packages/types/src/providers/poe.ts
@ -0,0 +1,2 @@
+export { poeDefaultModelId, POE_DEFAULT_BASE_URL, getPoeDefaultModelInfo } from "ai-sdk-provider-poe/code"
+export type { PoeDefaultModelInfo } from "ai-sdk-provider-poe/code"
--- a/packages/types/src/providers/xai.ts
+++ b/packages/types/src/providers/xai.ts
@ -3,10 +3,10 @@ import type { ModelInfo } from "../model.js"
 // https://docs.x.ai/docs/api-reference
 export type XAIModelId = keyof typeof xaiModels

-export const xaiDefaultModelId: XAIModelId = "grok-4.20-beta-0309-reasoning"
+export const xaiDefaultModelId: XAIModelId = "grok-4.20"

 export const xaiModels = {
-	"grok-4.20-beta-0309-reasoning": {
+	"grok-4.20": {
 		maxTokens: 65_536,
 		contextWindow: 2_000_000,
 		supportsImages: true,
@ -15,21 +15,7 @@ export const xaiModels = {
 		outputPrice: 6.0,
 		cacheWritesPrice: 0.5,
 		cacheReadsPrice: 0.5,
-		description:
-			"xAI's Grok 4.20 reasoning model with 2M context. Reasoning is internal (not exposed via Chat Completions API).",
-		includedTools: ["search_replace"],
-		excludedTools: ["apply_diff"],
-	},
-	"grok-4.20-beta-0309-non-reasoning": {
-		maxTokens: 65_536,
-		contextWindow: 2_000_000,
-		supportsImages: true,
-		supportsPromptCache: true,
-		inputPrice: 2.0,
-		outputPrice: 6.0,
-		cacheWritesPrice: 0.5,
-		cacheReadsPrice: 0.5,
-		description: "xAI's Grok 4.20 non-reasoning model - faster inference with 2M context.",
+		description: "xAI's flagship Grok 4.20 model with 2M context and reasoning support via Responses API.",
 		includedTools: ["search_replace"],
 		excludedTools: ["apply_diff"],
 	},
--- a/packages/types/src/roomodes-schema.ts
+++ b/packages/types/src/roomodes-schema.ts
@ -0,0 +1,58 @@
+/**
+ * Builds the Zod schema for .roomodes configuration files and converts it
+ * to JSON Schema (draft-07). This module is the single source of truth for
+ * both the generator script (scripts/generate-roomodes-schema.ts) and the
+ * drift-detection test.
+ */
+
+import { z } from "zod"
+import { zodToJsonSchema } from "zod-to-json-schema"
+
+import { toolGroups, deprecatedToolGroups } from "./tool.js"
+import { groupOptionsSchema, modeConfigSchema } from "./mode.js"
+
+// Build a ToolGroup enum that includes deprecated groups so existing configs
+// still validate.
+const allToolGroups = [...toolGroups, ...deprecatedToolGroups] as [string, ...string[]]
+const allToolGroupsSchema = z.enum(allToolGroups)
+
+// Build a GroupEntry schema that uses the extended tool group list.
+const groupEntrySchema = z.union([allToolGroupsSchema, z.tuple([allToolGroupsSchema, groupOptionsSchema])])
+
+// Build the RuleFile schema (used during import/export but not part of the
+// core Zod types).
+const ruleFileSchema = z.object({
+	relativePath: z.string(),
+	content: z.string().optional(),
+})
+
+// Build an extended ModeConfig schema that includes rulesFiles and uses the
+// extended groups (with deprecated entries).
+const exportedModeConfigSchema = modeConfigSchema.omit({ groups: true }).extend({
+	groups: z.array(groupEntrySchema),
+	rulesFiles: z.array(ruleFileSchema).optional(),
+})
+
+// Build the top-level .roomodes schema.
+const roomodesZodSchema = z
+	.object({
+		customModes: z.array(exportedModeConfigSchema),
+	})
+	.strict()
+
+/**
+ * Generates the JSON Schema object for .roomodes configuration files.
+ * Includes metadata fields ($id, title, description).
+ */
+export function generateRoomodesJsonSchema(): Record<string, unknown> {
+	const jsonSchema = zodToJsonSchema(roomodesZodSchema, {
+		$refStrategy: "none",
+		target: "jsonSchema7",
+	}) as Record<string, unknown>
+
+	jsonSchema["$id"] = "https://github.com/RooCodeInc/Roo-Code/blob/main/schemas/roomodes.json"
+	jsonSchema["title"] = "Roo Code Custom Modes"
+	jsonSchema["description"] = "Schema for .roomodes configuration files used by Roo Code to define custom modes."
+
+	return jsonSchema
+}
--- a/pnpm-lock.yaml
+++ b/pnpm-lock.yaml
@ -350,8 +350,8 @@ importers:
        specifier: workspace:^
        version: link:../../packages/evals
      '@roo-code/types':
-        specifier: ^1.108.0
-        version: 1.108.0
+        specifier: workspace:^
+        version: link:../../packages/types
      '@tanstack/react-query':
        specifier: ^5.90.20
        version: 5.90.20(react@18.3.1)
@ -569,7 +569,7 @@ importers:
        version: 7.0.5
      openai:
        specifier: ^5.12.2
-        version: 5.12.2(ws@8.18.3)(zod@3.25.76)
+        version: 5.12.2(ws@8.20.0)(zod@3.25.76)
      zod:
        specifier: 3.25.76
        version: 3.25.76
@ -709,6 +709,9 @@ importers:

  packages/types:
    dependencies:
+      ai-sdk-provider-poe:
+        specifier: 2.0.18
+        version: 2.0.18(ai@6.0.77(zod@3.25.76))(zod@3.25.76)
      zod:
        specifier: 3.25.76
        version: 3.25.76
@ -722,6 +725,9 @@ importers:
      '@types/node':
        specifier: ^24.1.0
        version: 24.2.1
+      ajv:
+        specifier: ^8.18.0
+        version: 8.18.0
      globals:
        specifier: ^16.3.0
        version: 16.3.0
@ -731,6 +737,9 @@ importers:
      vitest:
        specifier: ^3.2.3
        version: 3.2.4(@types/debug@4.1.12)(@types/node@24.2.1)(@vitest/ui@3.2.4)(jiti@2.4.2)(jsdom@26.1.0)(lightningcss@1.30.1)(tsx@4.19.4)(yaml@2.8.0)
+      zod-to-json-schema:
+        specifier: ^3.25.1
+        version: 3.25.1(zod@3.25.76)

  packages/vscode-shim:
    devDependencies:
@ -818,6 +827,9 @@ importers:
      '@vscode/codicons':
        specifier: ^0.0.36
        version: 0.0.36
+      ai-sdk-provider-poe:
+        specifier: 2.0.18
+        version: 2.0.18(ai@6.0.77(zod@3.25.76))(zod@3.25.76)
      async-mutex:
        specifier: ^0.5.0
        version: 0.5.0
@ -907,7 +919,7 @@ importers:
        version: 0.5.17
      openai:
        specifier: ^5.12.2
-        version: 5.12.2(ws@8.18.3)(zod@3.25.76)
+        version: 5.12.2(ws@8.20.0)(zod@3.25.76)
      os-name:
        specifier: ^6.0.0
        version: 6.1.0
@ -1426,6 +1438,12 @@ packages:
    peerDependencies:
      zod: 3.25.76

+  '@ai-sdk/anthropic@3.0.42':
+    resolution: {integrity: sha512-snoLXB9DmvAmmngbPN/Io8IGzZ9zWpC208EgIIztYf1e1JhwuMkgKCYkL30vGhSen4PrBafu2+sO4G/17wu45A==}
+    engines: {node: '>=18'}
+    peerDependencies:
+      zod: 3.25.76
+
  '@ai-sdk/baseten@1.0.31':
    resolution: {integrity: sha512-tGbV96WBb5nnfyUYFrPyBxrhw53YlKSJbMC+rH3HhQlUaIs8+m/Bm4M0isrek9owIIf4MmmSDZ5VZL08zz7eFQ==}
    engines: {node: '>=18'}
@ -1480,6 +1498,18 @@ packages:
    peerDependencies:
      zod: 3.25.76

+  '@ai-sdk/openai-compatible@2.0.37':
+    resolution: {integrity: sha512-+POSFVcgiu47BK64dhsI6OpcDC0/VAE2ZSaXdXGNNhpC/ava++uSRJYks0k2bpfY0wwCTgpAWZsXn/dG2Yppiw==}
+    engines: {node: '>=18'}
+    peerDependencies:
+      zod: 3.25.76
+
+  '@ai-sdk/openai@3.0.27':
+    resolution: {integrity: sha512-pLMxWOypwroXiK9dxNpn60/HGhWWWDEOJ3lo9vZLoxvpJNtKnLKojwVIvlW3yEjlD7ll1+jUO2uzsABNTaP5Yg==}
+    engines: {node: '>=18'}
+    peerDependencies:
+      zod: 3.25.76
+
  '@ai-sdk/provider-utils@3.0.5':
    resolution: {integrity: sha512-HliwB/yzufw3iwczbFVE2Fiwf1XqROB/I6ng8EKUsPM5+2wnIa8f4VbljZcDx+grhFrPV+PnRZH7zBqi8WZM7Q==}
    engines: {node: '>=18'}
@ -1492,6 +1522,12 @@ packages:
    peerDependencies:
      zod: 3.25.76

+  '@ai-sdk/provider-utils@4.0.21':
+    resolution: {integrity: sha512-MtFUYI1/8mgDvRmaBDjbLJPFFrMG777AvSgyIFQtZHIMzm88R/12vYBBpnk7pfiWLFE1DSZzY4WDYzGbKAcmiw==}
+    engines: {node: '>=18'}
+    peerDependencies:
+      zod: 3.25.76
+
  '@ai-sdk/provider@2.0.0':
    resolution: {integrity: sha512-6o7Y2SeO9vFKB8lArHXehNuusnpddKPk7xqL7T2/b+OvXMRIXUO1rR4wcv1hAFUAT9avGZshty3Wlua/XA7TvA==}
    engines: {node: '>=18'}
@ -1726,6 +1762,10 @@ packages:
    resolution: {integrity: sha512-cjQ7ZlQ0Mv3b47hABuTevyTuYN4i+loJKGeV9flcCgIK37cCXRh+L1bd3iBHlynerhQ7BhCkn2BPbQUL+rGqFg==}
    engines: {node: '>=6.9.0'}

+  '@babel/code-frame@7.29.0':
+    resolution: {integrity: sha512-9NhCeYjq9+3uxgdtp20LSiJXJvN0FeCtNGpJxuMFZ1Kv3cWUNb6DOhJwUvcVCzKGR66cw4njwM6hrJLqgOwbcw==}
+    engines: {node: '>=6.9.0'}
+
  '@babel/compat-data@7.27.2':
    resolution: {integrity: sha512-TUtMJYRPyUb/9aU8f3K0mjmjf6M9N5Woshn2CS6nqJSeJtTtQcpLUXjGt9vbF8ZGff0El99sWkLgzwW3VXnxZQ==}
    engines: {node: '>=6.9.0'}
@ -1764,6 +1804,10 @@ packages:
    resolution: {integrity: sha512-D2hP9eA+Sqx1kBZgzxZh0y1trbuU+JoDkiEwqhQ36nodYqJwyEIhPSdMNd7lOm/4io72luTPWH20Yda0xOuUow==}
    engines: {node: '>=6.9.0'}

+  '@babel/helper-validator-identifier@7.28.5':
+    resolution: {integrity: sha512-qSs4ifwzKJSV39ucNjsvc6WVHs6b7S03sOh2OcHF9UHfVPqWWALUsNUVzhSBiItjRZoLHx7nIarVjqKVusUZ1Q==}
+    engines: {node: '>=6.9.0'}
+
  '@babel/helper-validator-option@7.27.1':
    resolution: {integrity: sha512-YvjJow9FxbhFFKDSuFnVCe2WxXk1zWc22fFePVNEaWJEu8IrZVlda6N0uHwzZrUM1il7NC9Mlp4MaJYbYd9JSg==}
    engines: {node: '>=6.9.0'}
@ -1805,6 +1849,10 @@ packages:
    resolution: {integrity: sha512-Q/N6JNWvIvPnLDvjlE1OUBLPQHH6l3CltCEsHIujp45zQUSSh8K+gHnaEX45yAT1nyngnINhvWtzN+Nb9D8RAQ==}
    engines: {node: '>=6.9.0'}

+  '@babel/runtime@7.29.2':
+    resolution: {integrity: sha512-JiDShH45zKHWyGe4ZNVRrCjBz8Nh9TMmZG1kh4QTK8hCBTWBi8Da+i7s1fJw7/lYpM4ccepSNfqzZ/QvABBi5g==}
+    engines: {node: '>=6.9.0'}
+
  '@babel/template@7.27.2':
    resolution: {integrity: sha512-LPDZ85aEJyYSd18/DkjNh4/y1ntkE5KwUHWTiqgRxruuZL2F1yuHligVHLvcHY2vMHXttKFpJn6LwfI7cw7ODw==}
    engines: {node: '>=6.9.0'}
@ -2532,13 +2580,13 @@ packages:
  '@libsql/core@0.15.15':
    resolution: {integrity: sha512-C88Z6UKl+OyuKKPwz224riz02ih/zHYI3Ho/LAcVOgjsunIRZoBw7fjRfaH9oPMmSNeQfhGklSG2il1URoOIsA==}

-  '@libsql/darwin-arm64@0.5.22':
-    resolution: {integrity: sha512-4B8ZlX3nIDPndfct7GNe0nI3Yw6ibocEicWdC4fvQbSs/jdq/RC2oCsoJxJ4NzXkvktX70C1J4FcmmoBy069UA==}
+  '@libsql/darwin-arm64@0.5.29':
+    resolution: {integrity: sha512-K+2RIB1OGFPYQbfay48GakLhqf3ArcbHqPFu7EZiaUcRgFcdw8RoltsMyvbj5ix2fY0HV3Q3Ioa/ByvQdaSM0A==}
    cpu: [arm64]
    os: [darwin]

-  '@libsql/darwin-x64@0.5.22':
-    resolution: {integrity: sha512-ny2HYWt6lFSIdNFzUFIJ04uiW6finXfMNJ7wypkAD8Pqdm6nAByO+Fdqu8t7sD0sqJGeUCiOg480icjyQ2/8VA==}
+  '@libsql/darwin-x64@0.5.29':
+    resolution: {integrity: sha512-OtT+KFHsKFy1R5FVadr8FJ2Bb1mghtXTyJkxv0trocq7NuHntSki1eUbxpO5ezJesDvBlqFjnWaYYY516QNLhQ==}
    cpu: [x64]
    os: [darwin]

@ -2552,38 +2600,38 @@ packages:
  '@libsql/isomorphic-ws@0.1.5':
    resolution: {integrity: sha512-DtLWIH29onUYR00i0GlQ3UdcTRC6EP4u9w/h9LxpUZJWRMARk6dQwZ6Jkd+QdwVpuAOrdxt18v0K2uIYR3fwFg==}

-  '@libsql/linux-arm-gnueabihf@0.5.22':
-    resolution: {integrity: sha512-3Uo3SoDPJe/zBnyZKosziRGtszXaEtv57raWrZIahtQDsjxBVjuzYQinCm9LRCJCUT5t2r5Z5nLDPJi2CwZVoA==}
+  '@libsql/linux-arm-gnueabihf@0.5.29':
+    resolution: {integrity: sha512-CD4n4zj7SJTHso4nf5cuMoWoMSS7asn5hHygsDuhRl8jjjCTT3yE+xdUvI4J7zsyb53VO5ISh4cwwOtf6k2UhQ==}
    cpu: [arm]
    os: [linux]

-  '@libsql/linux-arm-musleabihf@0.5.22':
-    resolution: {integrity: sha512-LCsXh07jvSojTNJptT9CowOzwITznD+YFGGW+1XxUr7fS+7/ydUrpDfsMX7UqTqjm7xG17eq86VkWJgHJfvpNg==}
+  '@libsql/linux-arm-musleabihf@0.5.29':
+    resolution: {integrity: sha512-2Z9qBVpEJV7OeflzIR3+l5yAd4uTOLxklScYTwpZnkm2vDSGlC1PRlueLaufc4EFITkLKXK2MWBpexuNJfMVcg==}
    cpu: [arm]
    os: [linux]

-  '@libsql/linux-arm64-gnu@0.5.22':
-    resolution: {integrity: sha512-KSdnOMy88c9mpOFKUEzPskSaF3VLflfSUCBwas/pn1/sV3pEhtMF6H8VUCd2rsedwoukeeCSEONqX7LLnQwRMA==}
+  '@libsql/linux-arm64-gnu@0.5.29':
+    resolution: {integrity: sha512-gURBqaiXIGGwFNEaUj8Ldk7Hps4STtG+31aEidCk5evMMdtsdfL3HPCpvys+ZF/tkOs2MWlRWoSq7SOuCE9k3w==}
    cpu: [arm64]
    os: [linux]

-  '@libsql/linux-arm64-musl@0.5.22':
-    resolution: {integrity: sha512-mCHSMAsDTLK5YH//lcV3eFEgiR23Ym0U9oEvgZA0667gqRZg/2px+7LshDvErEKv2XZ8ixzw3p1IrBzLQHGSsw==}
+  '@libsql/linux-arm64-musl@0.5.29':
+    resolution: {integrity: sha512-fwgYZ0H8mUkyVqXZHF3mT/92iIh1N94Owi/f66cPVNsk9BdGKq5gVpoKO+7UxaNzuEH1roJp2QEwsCZMvBLpqg==}
    cpu: [arm64]
    os: [linux]

-  '@libsql/linux-x64-gnu@0.5.22':
-    resolution: {integrity: sha512-kNBHaIkSg78Y4BqAdgjcR2mBilZXs4HYkAmi58J+4GRwDQZh5fIUWbnQvB9f95DkWUIGVeenqLRFY2pcTmlsew==}
+  '@libsql/linux-x64-gnu@0.5.29':
+    resolution: {integrity: sha512-y14V0vY0nmMC6G0pHeJcEarcnGU2H6cm21ZceRkacWHvQAEhAG0latQkCtoS2njFOXiYIg+JYPfAoWKbi82rkg==}
    cpu: [x64]
    os: [linux]

-  '@libsql/linux-x64-musl@0.5.22':
-    resolution: {integrity: sha512-UZ4Xdxm4pu3pQXjvfJiyCzZop/9j/eA2JjmhMaAhe3EVLH2g11Fy4fwyUp9sT1QJYR1kpc2JLuybPM0kuXv/Tg==}
+  '@libsql/linux-x64-musl@0.5.29':
+    resolution: {integrity: sha512-gquqwA/39tH4pFl+J9n3SOMSymjX+6kZ3kWgY3b94nXFTwac9bnFNMffIomgvlFaC4ArVqMnOZD3nuJ3H3VO1w==}
    cpu: [x64]
    os: [linux]

-  '@libsql/win32-x64-msvc@0.5.22':
-    resolution: {integrity: sha512-Fj0j8RnBpo43tVZUVoNK6BV/9AtDUM5S7DF3LB4qTYg1LMSZqi3yeCneUTLJD6XomQJlZzbI4mst89yspVSAnA==}
+  '@libsql/win32-x64-msvc@0.5.29':
+    resolution: {integrity: sha512-4/0CvEdhi6+KjMxMaVbFM2n2Z44escBRoEYpR+gZg64DdetzGnYm8mcNLcoySaDJZNaBd6wz5DNdgRmcI4hXcg==}
    cpu: [x64]
    os: [win32]

@ -3850,9 +3898,6 @@ packages:
    cpu: [x64]
    os: [win32]

-  '@roo-code/types@1.108.0':
-    resolution: {integrity: sha512-0Of0DOuU125i1VTI2OTdL0j47xA+JrQr6KyYinYS+CPwUsszUJt2PeWyy/AZYI1w23FYrcCvh8FqycDuwXocSA==}
-
  '@sec-ant/readable-stream@0.4.1':
    resolution: {integrity: sha512-831qok9r2t8AlxLko40y2ebgSDhenenCatLVeW/uBtnHPyhHOvG0C7TvfgecV+wHzIm5KUICgzmVpWS+IMEAeg==}

@ -4886,6 +4931,11 @@ packages:
    resolution: {integrity: sha512-kja8j7PjmncONqaTsB8fQ+wE2mSU2DJ9D4XKoJ5PFWIdRMa6SLSN1ff4mOr4jCbfRSsxR4keIiySJU0N9T5hIQ==}
    engines: {node: '>= 8.0.0'}

+  ai-sdk-provider-poe@2.0.18:
+    resolution: {integrity: sha512-uzFR5Zq+PD6LfrqpYAr4dt2KmfOfS9a0Wh8QJsOaGI/vOByhW02P/0mj0NtOnOWF1RIUfBkQJo647XjrgnFjjg==}
+    peerDependencies:
+      ai: '>=6.0.0'
+
  ai@6.0.77:
    resolution: {integrity: sha512-tyyhrRpCRFVlivdNIFLK8cexSBB2jwTqO0z1qJQagk+UxZ+MW8h5V8xsvvb+xdKDY482Y8KAm0mr7TDnPKvvlw==}
    engines: {node: '>=18'}
@ -4895,6 +4945,9 @@ packages:
  ajv@6.12.6:
    resolution: {integrity: sha512-j3fVLgvTo527anyYyJOGTYJbG+vnnQYvE0m5mmkc1TK+nxAppkCLMIL0aZ4dblVCNoGShhm+kzE4ZUykBoMg4g==}

+  ajv@8.18.0:
+    resolution: {integrity: sha512-PlXPeEWMXMZ7sPYOHqmDyCJzcfNrUr3fGNKtezX14ykXOEIvyK81d+qydx89KY5O71FKMPaQ2vBfBFI5NHR63A==}
+
  ansi-colors@4.1.3:
    resolution: {integrity: sha512-/6w/C21Pm1A7aZitlI5Ni/2J6FFQN8i1Cvz3kHABAAbw93v/NlvKdVOqz7CCWz/3iv/JplRSEEZ83XION15ovw==}
    engines: {node: '>=6'}
@ -5130,6 +5183,7 @@ packages:
  basic-ftp@5.0.5:
    resolution: {integrity: sha512-4Bcg1P8xhUuqcii/S0Z9wiHIrQVPMermM1any+MX5GeGD7faD3/msQUDGLol9wOcz4/jbg/WJnGqoJF6LiBdtg==}
    engines: {node: '>=10.0.0'}
+    deprecated: Security vulnerability fixed in 5.2.0, please upgrade

  better-path-resolve@1.0.0:
    resolution: {integrity: sha512-pbnl5XzGBdrFU/wT4jqmJVPn2B6UHPBOhzMQkY/SPUPB6QtUXtmBHBIwCbXJol93mOpGMnQyP/+BB19q04xj7g==}
@ -6577,6 +6631,9 @@ packages:
  fast-shallow-equal@1.0.0:
    resolution: {integrity: sha512-HPtaa38cPgWvaCFmRNhlc6NG7pv6NUHqjPgVAkWGoB9mQMwYB27/K0CvOM5Czy+qpT3e8XJ6Q4aPAnzpNpzNaw==}

+  fast-uri@3.1.0:
+    resolution: {integrity: sha512-iPeeDKJSWf4IEOasVVrknXpaBV0IApz/gp7S2bb7Z4Lljbl2MGJRqInZiUrQwV16cpzw/D3S5j5Julj/gT52AA==}
+
  fast-xml-parser@5.2.3:
    resolution: {integrity: sha512-OdCYfRqfpuLUFonTNjvd30rCBZUneHpSQkCqfaeWQ9qrKcl6XlWeDBNVwGb+INAIxRshuN2jF+BE0L6gbBO2mw==}
    hasBin: true
@ -7605,6 +7662,9 @@ packages:
  json-schema-traverse@0.4.1:
    resolution: {integrity: sha512-xbbCH5dCYU5T8LcEhhuh7HJ88HXuW3qsI3Y0zOZFKfZEHcpWiHU/Jxzk629Brsab/mMiHQti9wMP+845RPe3Vg==}

+  json-schema-traverse@1.0.0:
+    resolution: {integrity: sha512-NM8/P9n3XjXhIZn1lLhkFaACTOURQXjWhV4BA/RnOv8xvgqtqpAX9IO4mRQxSx1Rlo4tqzeqb0sOlruaOy3dug==}
+
  json-schema@0.4.0:
    resolution: {integrity: sha512-es94M3nTIfsEPisRafak+HDLfHXnKBhV3vU5eqPcS3flIWqcxJWgXHXiey3YrpaNsanY5ei1VoYEbOzijuq9BA==}

@ -7712,8 +7772,8 @@ packages:
    resolution: {integrity: sha512-+bT2uH4E5LGE7h/n3evcS/sQlJXCpIp6ym8OWJ5eV6+67Dsql/LaaT7qJBAt2rzfoa/5QBGBhxDix1dMt2kQKQ==}
    engines: {node: '>= 0.8.0'}

-  libsql@0.5.22:
-    resolution: {integrity: sha512-NscWthMQt7fpU8lqd7LXMvT9pi+KhhmTHAJWUB/Lj6MWa0MKFv0F2V4C6WKKpjCVZl0VwcDz4nOI3CyaT1DDiA==}
+  libsql@0.5.29:
+    resolution: {integrity: sha512-8lMP8iMgiBzzoNbAPQ59qdVcj6UaE/Vnm+fiwX4doX4Narook0a4GPKWBEv+CR8a1OwbfkgL18uBfBjWdF0Fzg==}
    cpu: [x64, arm64, wasm32, arm]
    os: [darwin, linux, win32]

@ -8976,6 +9036,7 @@ packages:
  prebuild-install@7.1.3:
    resolution: {integrity: sha512-8Mf2cbV7x1cXPUILADGI3wuhfqWvtiLA1iclTDbFRZkgRQS0NqsPZphna9V+HyTEadheuPmjaJMsbzKQFOzLug==}
    engines: {node: '>=10'}
+    deprecated: No longer maintained. Please contact the author of the relevant native addon; alternatives are available.
    hasBin: true

  prelude-ls@1.2.1:
@ -9386,6 +9447,10 @@ packages:
    resolution: {integrity: sha512-fGxEI7+wsG9xrvdjsrlmL22OMTTiHRwAMroiEeMgq8gzoLC/PQr7RsRDSTLUg/bZAZtF+TVIkHc6/4RIKrui+Q==}
    engines: {node: '>=0.10.0'}

+  require-from-string@2.0.2:
+    resolution: {integrity: sha512-Xf0nWe6RseziFMu+Ap9biiUbmplq6S9/p+7w7YXP/JBHhrUDDUhwa+vANyubuqfZWTveU//DYVGsDG7RKL/vEw==}
+    engines: {node: '>=0.10.0'}
+
  resize-observer-polyfill@1.5.1:
    resolution: {integrity: sha512-LwZrotdHOo12nQuZlHEmtuXdqGoOD0OhaxopaNFxWzInpEgaLWoVuAMbTzixuosCx2nEG58ngzW3vxdWoxIgdg==}

@ -9565,6 +9630,11 @@ packages:
    engines: {node: '>=10'}
    hasBin: true

+  semver@7.7.4:
+    resolution: {integrity: sha512-vFKC2IEtQnVhpT78h1Yp8wzwrf8CM+MzKMHGJZfBtzhZNycRFnXsHk6E5TxIkkMsgNS7mdX3AGB7x2QM2di4lA==}
+    engines: {node: '>=10'}
+    hasBin: true
+
  send@1.2.0:
    resolution: {integrity: sha512-uaW0WwXKpL9blXE2o0bRhoL2EGXIrZxQ2ZQ4mgcfoBxdFmQold+qWsD2jLrfZ0trjKL6vOw0j//eAwcALFjKSw==}
    engines: {node: '>= 18'}
@ -10898,6 +10968,18 @@ packages:
      utf-8-validate:
        optional: true

+  ws@8.20.0:
+    resolution: {integrity: sha512-sAt8BhgNbzCtgGbt2OxmpuryO63ZoDk/sqaB/znQm94T4fCEsy/yV+7CdC1kJhOU9lboAEU7R3kquuycDoibVA==}
+    engines: {node: '>=10.0.0'}
+    peerDependencies:
+      bufferutil: ^4.0.1
+      utf-8-validate: '>=5.0.2'
+    peerDependenciesMeta:
+      bufferutil:
+        optional: true
+      utf-8-validate:
+        optional: true
+
  xml-name-validator@5.0.0:
    resolution: {integrity: sha512-EvGK8EJ3DhaHfbRlETOWAS5pO9MZITeauHKJyb8wyajUfQUenkIg2MvLDTZ4T/TgIcm3HU0TFBgWWboAZ30UHg==}
    engines: {node: '>=18'}
@ -11002,6 +11084,11 @@ packages:
    peerDependencies:
      zod: 3.25.76

+  zod-to-json-schema@3.25.1:
+    resolution: {integrity: sha512-pM/SU9d3YAggzi6MtR4h7ruuQlqKtad8e9S0fmxcMi+ueAK5Korys/aWcV9LIIHTVbj01NdzxcnXSN+O74ZIVA==}
+    peerDependencies:
+      zod: 3.25.76
+
  zod-to-ts@1.2.0:
    resolution: {integrity: sha512-x30XE43V+InwGpvTySRNz9kB7qFU8DlyEy7BsSTCHPH1R0QasMmHWZDCzYm6bVXtj/9NNJAZF3jW8rzFvH5OFA==}
    peerDependencies:
@ -11058,6 +11145,12 @@ snapshots:
      '@ai-sdk/provider-utils': 4.0.14(zod@3.25.76)
      zod: 3.25.76

+  '@ai-sdk/anthropic@3.0.42(zod@3.25.76)':
+    dependencies:
+      '@ai-sdk/provider': 3.0.8
+      '@ai-sdk/provider-utils': 4.0.14(zod@3.25.76)
+      zod: 3.25.76
+
  '@ai-sdk/baseten@1.0.31(zod@3.25.76)':
    dependencies:
      '@ai-sdk/openai-compatible': 2.0.28(zod@3.25.76)
@ -11121,6 +11214,18 @@ snapshots:
      '@ai-sdk/provider-utils': 4.0.14(zod@3.25.76)
      zod: 3.25.76

+  '@ai-sdk/openai-compatible@2.0.37(zod@3.25.76)':
+    dependencies:
+      '@ai-sdk/provider': 3.0.8
+      '@ai-sdk/provider-utils': 4.0.21(zod@3.25.76)
+      zod: 3.25.76
+
+  '@ai-sdk/openai@3.0.27(zod@3.25.76)':
+    dependencies:
+      '@ai-sdk/provider': 3.0.8
+      '@ai-sdk/provider-utils': 4.0.14(zod@3.25.76)
+      zod: 3.25.76
+
  '@ai-sdk/provider-utils@3.0.5(zod@3.25.76)':
    dependencies:
      '@ai-sdk/provider': 2.0.0
@ -11136,6 +11241,13 @@ snapshots:
      eventsource-parser: 3.0.6
      zod: 3.25.76

+  '@ai-sdk/provider-utils@4.0.21(zod@3.25.76)':
+    dependencies:
+      '@ai-sdk/provider': 3.0.8
+      '@standard-schema/spec': 1.1.0
+      eventsource-parser: 3.0.6
+      zod: 3.25.76
+
  '@ai-sdk/provider@2.0.0':
    dependencies:
      json-schema: 0.4.0
@ -11767,6 +11879,12 @@ snapshots:
      js-tokens: 4.0.0
      picocolors: 1.1.1

+  '@babel/code-frame@7.29.0':
+    dependencies:
+      '@babel/helper-validator-identifier': 7.28.5
+      js-tokens: 4.0.0
+      picocolors: 1.1.1
+
  '@babel/compat-data@7.27.2': {}

  '@babel/core@7.27.1':
@ -11827,6 +11945,8 @@ snapshots:

  '@babel/helper-validator-identifier@7.27.1': {}

+  '@babel/helper-validator-identifier@7.28.5': {}
+
  '@babel/helper-validator-option@7.27.1': {}

  '@babel/helpers@7.27.1':
@ -11856,6 +11976,8 @@ snapshots:

  '@babel/runtime@7.28.4': {}

+  '@babel/runtime@7.29.2': {}
+
  '@babel/template@7.27.2':
    dependencies:
      '@babel/code-frame': 7.27.1
@ -12575,7 +12697,7 @@ snapshots:
      '@libsql/core': 0.15.15
      '@libsql/hrana-client': 0.7.0
      js-base64: 3.7.8
-      libsql: 0.5.22
+      libsql: 0.5.29
      promise-limit: 2.7.0
    transitivePeerDependencies:
      - bufferutil
@ -12587,10 +12709,10 @@ snapshots:
      js-base64: 3.7.8
    optional: true

-  '@libsql/darwin-arm64@0.5.22':
+  '@libsql/darwin-arm64@0.5.29':
    optional: true

-  '@libsql/darwin-x64@0.5.22':
+  '@libsql/darwin-x64@0.5.29':
    optional: true

  '@libsql/hrana-client@0.7.0':
@ -12610,31 +12732,31 @@ snapshots:
  '@libsql/isomorphic-ws@0.1.5':
    dependencies:
      '@types/ws': 8.18.1
-      ws: 8.18.3
+      ws: 8.20.0
    transitivePeerDependencies:
      - bufferutil
      - utf-8-validate
    optional: true

-  '@libsql/linux-arm-gnueabihf@0.5.22':
+  '@libsql/linux-arm-gnueabihf@0.5.29':
    optional: true

-  '@libsql/linux-arm-musleabihf@0.5.22':
+  '@libsql/linux-arm-musleabihf@0.5.29':
    optional: true

-  '@libsql/linux-arm64-gnu@0.5.22':
+  '@libsql/linux-arm64-gnu@0.5.29':
    optional: true

-  '@libsql/linux-arm64-musl@0.5.22':
+  '@libsql/linux-arm64-musl@0.5.29':
    optional: true

-  '@libsql/linux-x64-gnu@0.5.22':
+  '@libsql/linux-x64-gnu@0.5.29':
    optional: true

-  '@libsql/linux-x64-musl@0.5.22':
+  '@libsql/linux-x64-musl@0.5.29':
    optional: true

-  '@libsql/win32-x64-msvc@0.5.22':
+  '@libsql/win32-x64-msvc@0.5.29':
    optional: true

  '@lmstudio/lms-isomorphic@0.4.5':
@ -13863,10 +13985,6 @@ snapshots:
  '@rollup/rollup-win32-x64-msvc@4.40.2':
    optional: true

-  '@roo-code/types@1.108.0':
-    dependencies:
-      zod: 3.25.76
-
  '@sec-ant/readable-stream@0.4.1': {}

  '@sevinf/maybe@0.5.0': {}
@ -14389,8 +14507,8 @@ snapshots:

  '@testing-library/dom@10.4.0':
    dependencies:
-      '@babel/code-frame': 7.27.1
-      '@babel/runtime': 7.28.4
+      '@babel/code-frame': 7.29.0
+      '@babel/runtime': 7.29.2
      '@types/aria-query': 5.0.4
      aria-query: 5.3.0
      chalk: 4.1.2
@ -15129,6 +15247,17 @@ snapshots:
    dependencies:
      humanize-ms: 1.2.1

+  ai-sdk-provider-poe@2.0.18(ai@6.0.77(zod@3.25.76))(zod@3.25.76):
+    dependencies:
+      '@ai-sdk/anthropic': 3.0.42(zod@3.25.76)
+      '@ai-sdk/openai': 3.0.27(zod@3.25.76)
+      '@ai-sdk/openai-compatible': 2.0.37(zod@3.25.76)
+      '@ai-sdk/provider': 3.0.8
+      '@ai-sdk/provider-utils': 4.0.14(zod@3.25.76)
+      ai: 6.0.77(zod@3.25.76)
+    transitivePeerDependencies:
+      - zod
+
  ai@6.0.77(zod@3.25.76):
    dependencies:
      '@ai-sdk/gateway': 3.0.39(zod@3.25.76)
@ -15144,6 +15273,13 @@ snapshots:
      json-schema-traverse: 0.4.1
      uri-js: 4.4.1

+  ajv@8.18.0:
+    dependencies:
+      fast-deep-equal: 3.1.3
+      fast-uri: 3.1.0
+      json-schema-traverse: 1.0.0
+      require-from-string: 2.0.2
+
  ansi-colors@4.1.3: {}

  ansi-escapes@7.0.0:
@ -16986,6 +17122,8 @@ snapshots:

  fast-shallow-equal@1.0.0: {}

+  fast-uri@3.1.0: {}
+
  fast-xml-parser@5.2.3:
    dependencies:
      strnum: 2.1.1
@ -17232,7 +17370,7 @@ snapshots:
      '@petamoriken/float16': 3.9.3
      debug: 4.4.3
      env-paths: 3.0.0
-      semver: 7.7.3
+      semver: 7.7.4
      shell-quote: 1.8.3
      which: 4.0.0
    transitivePeerDependencies:
@ -18115,6 +18253,8 @@ snapshots:

  json-schema-traverse@0.4.1: {}

+  json-schema-traverse@1.0.0: {}
+
  json-schema@0.4.0: {}

  json-stable-stringify-without-jsonify@1.0.1: {}
@ -18251,20 +18391,20 @@ snapshots:
      prelude-ls: 1.2.1
      type-check: 0.4.0

-  libsql@0.5.22:
+  libsql@0.5.29:
    dependencies:
      '@neon-rs/load': 0.0.4
      detect-libc: 2.0.2
    optionalDependencies:
-      '@libsql/darwin-arm64': 0.5.22
-      '@libsql/darwin-x64': 0.5.22
-      '@libsql/linux-arm-gnueabihf': 0.5.22
-      '@libsql/linux-arm-musleabihf': 0.5.22
-      '@libsql/linux-arm64-gnu': 0.5.22
-      '@libsql/linux-arm64-musl': 0.5.22
-      '@libsql/linux-x64-gnu': 0.5.22
-      '@libsql/linux-x64-musl': 0.5.22
-      '@libsql/win32-x64-msvc': 0.5.22
+      '@libsql/darwin-arm64': 0.5.29
+      '@libsql/darwin-x64': 0.5.29
+      '@libsql/linux-arm-gnueabihf': 0.5.29
+      '@libsql/linux-arm-musleabihf': 0.5.29
+      '@libsql/linux-arm64-gnu': 0.5.29
+      '@libsql/linux-arm64-musl': 0.5.29
+      '@libsql/linux-x64-gnu': 0.5.29
+      '@libsql/linux-x64-musl': 0.5.29
+      '@libsql/win32-x64-msvc': 0.5.29
    optional: true

  lie@3.3.0:
@ -19384,9 +19524,9 @@ snapshots:
      is-inside-container: 1.0.0
      is-wsl: 3.1.0

-  openai@5.12.2(ws@8.18.3)(zod@3.25.76):
+  openai@5.12.2(ws@8.20.0)(zod@3.25.76):
    optionalDependencies:
-      ws: 8.18.3
+      ws: 8.20.0
      zod: 3.25.76

  option@0.2.4: {}
@ -20328,6 +20468,8 @@ snapshots:

  require-directory@2.1.1: {}

+  require-from-string@2.0.2: {}
+
  resize-observer-polyfill@1.5.1: {}

  resolve-from@4.0.0: {}
@ -20538,6 +20680,9 @@ snapshots:

  semver@7.7.3: {}

+  semver@7.7.4:
+    optional: true
+
  send@1.2.0:
    dependencies:
      debug: 4.4.1(supports-color@8.1.1)
@ -22165,6 +22310,9 @@ snapshots:

  ws@8.18.3: {}

+  ws@8.20.0:
+    optional: true
+
  xml-name-validator@5.0.0: {}

  xml2js@0.5.0:
@ -22267,6 +22415,10 @@ snapshots:
    dependencies:
      zod: 3.25.76

+  zod-to-json-schema@3.25.1(zod@3.25.76):
+    dependencies:
+      zod: 3.25.76
+
  zod-to-ts@1.2.0(typescript@5.8.3)(zod@3.25.76):
    dependencies:
      typescript: 5.8.3
--- a/schemas/roomodes.json
+++ b/schemas/roomodes.json
@ -0,0 +1,96 @@
+{
+	"type": "object",
+	"properties": {
+		"customModes": {
+			"type": "array",
+			"items": {
+				"type": "object",
+				"properties": {
+					"slug": {
+						"type": "string",
+						"pattern": "^[a-zA-Z0-9-]+$"
+					},
+					"name": {
+						"type": "string",
+						"minLength": 1
+					},
+					"roleDefinition": {
+						"type": "string",
+						"minLength": 1
+					},
+					"whenToUse": {
+						"type": "string"
+					},
+					"description": {
+						"type": "string"
+					},
+					"customInstructions": {
+						"type": "string"
+					},
+					"source": {
+						"type": "string",
+						"enum": ["global", "project"]
+					},
+					"groups": {
+						"type": "array",
+						"items": {
+							"anyOf": [
+								{
+									"type": "string",
+									"enum": ["read", "edit", "command", "mcp", "modes", "browser"]
+								},
+								{
+									"type": "array",
+									"minItems": 2,
+									"maxItems": 2,
+									"items": [
+										{
+											"type": "string",
+											"enum": ["read", "edit", "command", "mcp", "modes", "browser"]
+										},
+										{
+											"type": "object",
+											"properties": {
+												"fileRegex": {
+													"type": "string"
+												},
+												"description": {
+													"type": "string"
+												}
+											},
+											"additionalProperties": false
+										}
+									]
+								}
+							]
+						}
+					},
+					"rulesFiles": {
+						"type": "array",
+						"items": {
+							"type": "object",
+							"properties": {
+								"relativePath": {
+									"type": "string"
+								},
+								"content": {
+									"type": "string"
+								}
+							},
+							"required": ["relativePath"],
+							"additionalProperties": false
+						}
+					}
+				},
+				"required": ["slug", "name", "roleDefinition", "groups"],
+				"additionalProperties": false
+			}
+		}
+	},
+	"required": ["customModes"],
+	"additionalProperties": false,
+	"$schema": "http://json-schema.org/draft-07/schema#",
+	"$id": "https://github.com/RooCodeInc/Roo-Code/blob/main/schemas/roomodes.json",
+	"title": "Roo Code Custom Modes",
+	"description": "Schema for .roomodes configuration files used by Roo Code to define custom modes."
+}
--- a/src/api/index.ts
+++ b/src/api/index.ts
@ -9,6 +9,7 @@ import {
 	AnthropicHandler,
 	AwsBedrockHandler,
 	OpenRouterHandler,
+	PoeHandler,
 	VertexHandler,
 	AnthropicVertexHandler,
 	OpenAiHandler,
@ -176,6 +177,8 @@ export function buildApiHandler(configuration: ProviderSettings): ApiHandler {
 			return new MiniMaxHandler(options)
 		case "baseten":
 			return new BasetenHandler(options)
+		case "poe":
+			return new PoeHandler(options)
 		default:
 			return new AnthropicHandler(options)
 	}
--- a/src/api/providers/tests/minimax.spec.ts
+++ b/src/api/providers/tests/minimax.spec.ts
@ -112,7 +112,7 @@ describe("MiniMaxHandler", () => {
 			const model = handlerWithModel.getModel()
 			expect(model.id).toBe(testModelId)
 			expect(model.info).toEqual(minimaxModels[testModelId])
-			expect(model.info.contextWindow).toBe(192_000)
+			expect(model.info.contextWindow).toBe(204_800)
 			expect(model.info.maxTokens).toBe(16_384)
 			expect(model.info.supportsPromptCache).toBe(true)
 			expect(model.info.cacheWritesPrice).toBe(0.375)
@ -128,7 +128,7 @@ describe("MiniMaxHandler", () => {
 			const model = handlerWithModel.getModel()
 			expect(model.id).toBe(testModelId)
 			expect(model.info).toEqual(minimaxModels[testModelId])
-			expect(model.info.contextWindow).toBe(192_000)
+			expect(model.info.contextWindow).toBe(204_800)
 			expect(model.info.maxTokens).toBe(16_384)
 			expect(model.info.supportsPromptCache).toBe(true)
 			expect(model.info.cacheWritesPrice).toBe(0.375)
@ -191,10 +191,10 @@ describe("MiniMaxHandler", () => {
 			expect(model.info).toEqual(minimaxModels[minimaxDefaultModelId])
 		})

-		it("should default to MiniMax-M2.5 model", () => {
+		it("should default to MiniMax-M2.7 model", () => {
 			const handlerDefault = new MiniMaxHandler({ minimaxApiKey: "test-minimax-api-key" })
 			const model = handlerDefault.getModel()
-			expect(model.id).toBe("MiniMax-M2.5")
+			expect(model.id).toBe("MiniMax-M2.7")
 		})
 	})

@ -399,7 +399,7 @@ describe("MiniMaxHandler", () => {
 		it("should correctly configure MiniMax-M2 model properties", () => {
 			const model = minimaxModels["MiniMax-M2"]
 			expect(model.maxTokens).toBe(16_384)
-			expect(model.contextWindow).toBe(192_000)
+			expect(model.contextWindow).toBe(204_800)
 			expect(model.supportsImages).toBe(false)
 			expect(model.supportsPromptCache).toBe(true)
 			expect(model.inputPrice).toBe(0.3)
@ -411,7 +411,7 @@ describe("MiniMaxHandler", () => {
 		it("should correctly configure MiniMax-M2-Stable model properties", () => {
 			const model = minimaxModels["MiniMax-M2-Stable"]
 			expect(model.maxTokens).toBe(16_384)
-			expect(model.contextWindow).toBe(192_000)
+			expect(model.contextWindow).toBe(204_800)
 			expect(model.supportsImages).toBe(false)
 			expect(model.supportsPromptCache).toBe(true)
 			expect(model.inputPrice).toBe(0.3)
@ -419,5 +419,63 @@ describe("MiniMaxHandler", () => {
 			expect(model.cacheWritesPrice).toBe(0.375)
 			expect(model.cacheReadsPrice).toBe(0.03)
 		})
+
+		it("should correctly configure MiniMax-M2.7 model properties", () => {
+			const model = minimaxModels["MiniMax-M2.7"]
+			expect(model.maxTokens).toBe(16_384)
+			expect(model.contextWindow).toBe(204_800)
+			expect(model.supportsImages).toBe(false)
+			expect(model.supportsPromptCache).toBe(true)
+			expect(model.inputPrice).toBe(0.3)
+			expect(model.outputPrice).toBe(1.2)
+			expect(model.cacheWritesPrice).toBe(0.375)
+			expect(model.cacheReadsPrice).toBe(0.06)
+		})
+
+		it("should correctly configure MiniMax-M2.7-highspeed model properties", () => {
+			const model = minimaxModels["MiniMax-M2.7-highspeed"]
+			expect(model.maxTokens).toBe(16_384)
+			expect(model.contextWindow).toBe(204_800)
+			expect(model.supportsImages).toBe(false)
+			expect(model.supportsPromptCache).toBe(true)
+			expect(model.inputPrice).toBe(0.6)
+			expect(model.outputPrice).toBe(2.4)
+			expect(model.cacheWritesPrice).toBe(0.375)
+			expect(model.cacheReadsPrice).toBe(0.06)
+		})
+
+		it("should correctly configure MiniMax-M2.5-highspeed model properties", () => {
+			const model = minimaxModels["MiniMax-M2.5-highspeed"]
+			expect(model.maxTokens).toBe(16_384)
+			expect(model.contextWindow).toBe(204_800)
+			expect(model.supportsImages).toBe(false)
+			expect(model.supportsPromptCache).toBe(true)
+			expect(model.inputPrice).toBe(0.6)
+			expect(model.outputPrice).toBe(2.4)
+			expect(model.cacheWritesPrice).toBe(0.375)
+			expect(model.cacheReadsPrice).toBe(0.03)
+		})
+
+		it("should correctly configure MiniMax-M2.1-highspeed model properties", () => {
+			const model = minimaxModels["MiniMax-M2.1-highspeed"]
+			expect(model.maxTokens).toBe(16_384)
+			expect(model.contextWindow).toBe(204_800)
+			expect(model.supportsImages).toBe(false)
+			expect(model.supportsPromptCache).toBe(true)
+			expect(model.inputPrice).toBe(0.6)
+			expect(model.outputPrice).toBe(2.4)
+			expect(model.cacheWritesPrice).toBe(0.375)
+			expect(model.cacheReadsPrice).toBe(0.03)
+		})
+
+		it("should correctly configure MiniMax-M2.1 model properties with updated context window", () => {
+			const model = minimaxModels["MiniMax-M2.1"]
+			expect(model.contextWindow).toBe(204_800)
+		})
+
+		it("should correctly configure MiniMax-M2 model properties with updated context window", () => {
+			const model = minimaxModels["MiniMax-M2"]
+			expect(model.contextWindow).toBe(204_800)
+		})
 	})
 })
--- a/src/api/providers/tests/poe.spec.ts
+++ b/src/api/providers/tests/poe.spec.ts
@ -0,0 +1,307 @@
+const mockStreamText = vitest.fn()
+const mockGenerateText = vitest.fn()
+const mockCreatePoe = vitest.fn()
+
+vitest.mock("ai-sdk-provider-poe", () => ({
+	createPoe: (...args: unknown[]) => mockCreatePoe(...args),
+}))
+
+vitest.mock("ai-sdk-provider-poe/code", () => ({
+	mapToolChoice: vitest.fn((value: unknown) => value),
+	extractUsageMetrics: vitest.fn((usage: any) => ({
+		inputTokens: usage?.inputTokens || 0,
+		outputTokens: usage?.outputTokens || 0,
+		cacheReadTokens: usage?.cacheReadTokens,
+		cacheWriteTokens: usage?.cacheWriteTokens,
+		reasoningTokens: usage?.reasoningTokens,
+	})),
+	getPoeDefaultModelInfo: vitest.fn(() => ({
+		maxTokens: 8192,
+		contextWindow: 200_000,
+		supportsImages: true,
+		supportsPromptCache: true,
+		inputPrice: 3,
+		outputPrice: 15,
+	})),
+}))
+
+vitest.mock("ai", async (importOriginal) => {
+	const actual = await importOriginal<typeof import("ai")>()
+	return {
+		...actual,
+		streamText: (...args: unknown[]) => mockStreamText(...args),
+		generateText: (...args: unknown[]) => mockGenerateText(...args),
+	}
+})
+
+vitest.mock("../fetchers/modelCache", () => ({
+	getModelsFromCache: vitest.fn().mockReturnValue({
+		"anthropic/claude-sonnet-4": {
+			maxTokens: 10_000,
+			contextWindow: 200_000,
+			supportsImages: true,
+			supportsPromptCache: true,
+			supportsReasoningBudget: true,
+			inputPrice: 3,
+			outputPrice: 15,
+		},
+		"openai/gpt-4o": {
+			maxTokens: 16_384,
+			contextWindow: 128_000,
+			supportsImages: true,
+			supportsPromptCache: false,
+			inputPrice: 2.5,
+			outputPrice: 10,
+		},
+		"openai/o3": {
+			maxTokens: 100_000,
+			contextWindow: 200_000,
+			supportsImages: true,
+			supportsPromptCache: false,
+			supportsReasoningEffort: ["low", "medium", "high"],
+			inputPrice: 10,
+			outputPrice: 40,
+		},
+	}),
+}))
+
+import { poeDefaultModelId } from "@roo-code/types"
+import { PoeHandler } from "../poe"
+
+describe("PoeHandler", () => {
+	const mockLanguageModel = { modelId: "test-model" }
+	const mockPoeProvider = vitest.fn().mockReturnValue(mockLanguageModel)
+
+	beforeEach(() => {
+		vitest.clearAllMocks()
+		mockCreatePoe.mockReturnValue(mockPoeProvider)
+	})
+
+	describe("constructor", () => {
+		it("creates poe provider with api key and default base URL", () => {
+			new PoeHandler({ poeApiKey: "test-key" })
+
+			expect(mockCreatePoe).toHaveBeenCalledWith({
+				apiKey: "test-key",
+				baseURL: undefined,
+			})
+		})
+
+		it("creates poe provider with custom base URL", () => {
+			new PoeHandler({ poeApiKey: "key", poeBaseUrl: "https://custom.poe.com/v1" })
+
+			expect(mockCreatePoe).toHaveBeenCalledWith({
+				apiKey: "key",
+				baseURL: "https://custom.poe.com/v1",
+			})
+		})
+
+		it("uses fallback api key when not provided", () => {
+			new PoeHandler({})
+
+			expect(mockCreatePoe).toHaveBeenCalledWith({
+				apiKey: "not-provided",
+				baseURL: undefined,
+			})
+		})
+	})
+
+	describe("getModel", () => {
+		it("returns model info from cache", () => {
+			const handler = new PoeHandler({ poeApiKey: "key", apiModelId: "anthropic/claude-sonnet-4" })
+			const result = handler.getModel()
+
+			expect(result.id).toBe("anthropic/claude-sonnet-4")
+			expect(result.info.contextWindow).toBe(200_000)
+			expect(result.info.maxTokens).toBe(10_000)
+		})
+
+		it("returns default model when no model ID specified", () => {
+			const handler = new PoeHandler({ poeApiKey: "key" })
+			const result = handler.getModel()
+
+			expect(result.id).toBe(poeDefaultModelId)
+		})
+
+		it("falls back to default model info when model not in cache", () => {
+			const handler = new PoeHandler({ poeApiKey: "key", apiModelId: "unknown/model" })
+			const result = handler.getModel()
+
+			expect(result.id).toBe("unknown/model")
+			expect(result.info.contextWindow).toBeGreaterThan(0)
+			expect(result.info.maxTokens).toBeGreaterThan(0)
+		})
+	})
+
+	describe("createMessage", () => {
+		it("streams text chunks", async () => {
+			const handler = new PoeHandler({ poeApiKey: "key", apiModelId: "anthropic/claude-sonnet-4" })
+
+			const fullStream = (async function* () {
+				yield { type: "text-delta", text: "Hello " }
+				yield { type: "text-delta", text: "world!" }
+			})()
+
+			mockStreamText.mockReturnValue({
+				fullStream,
+				usage: Promise.resolve({ inputTokens: 10, outputTokens: 5 }),
+			})
+
+			const chunks = []
+			for await (const chunk of handler.createMessage("system prompt", [
+				{ role: "user" as const, content: "Hi" },
+			])) {
+				chunks.push(chunk)
+			}
+
+			expect(chunks).toContainEqual({ type: "text", text: "Hello " })
+			expect(chunks).toContainEqual({ type: "text", text: "world!" })
+			expect(chunks).toContainEqual(expect.objectContaining({ type: "usage", inputTokens: 10, outputTokens: 5 }))
+		})
+	})
+
+	describe("reasoning", () => {
+		it("passes anthropic thinking config for budget models", async () => {
+			const handler = new PoeHandler({
+				poeApiKey: "key",
+				apiModelId: "anthropic/claude-sonnet-4",
+				enableReasoningEffort: true,
+				modelMaxThinkingTokens: 4096,
+			})
+
+			const fullStream = (async function* () {
+				yield { type: "reasoning", text: "Let me think..." }
+				yield { type: "text-delta", text: "Answer" }
+			})()
+
+			mockStreamText.mockReturnValue({
+				fullStream,
+				usage: Promise.resolve({ inputTokens: 10, outputTokens: 5 }),
+			})
+
+			const chunks = []
+			const modelMaxTokens = handler.getModel().info.maxTokens ?? 0
+			for await (const chunk of handler.createMessage("system", [
+				{ role: "user" as const, content: "think about this" },
+			])) {
+				chunks.push(chunk)
+			}
+
+			const callArgs = mockStreamText.mock.calls[0][0]
+			expect(callArgs.temperature).toBe(1.0)
+			expect(callArgs.providerOptions).toEqual({
+				poe: {
+					reasoningBudgetTokens: 4096,
+				},
+			})
+			expect(callArgs.maxOutputTokens).toBe(modelMaxTokens - 4096)
+
+			expect(chunks).toContainEqual({ type: "reasoning", text: "Let me think..." })
+			expect(chunks).toContainEqual({ type: "text", text: "Answer" })
+		})
+
+		it("passes openai reasoning effort for effort models", async () => {
+			const handler = new PoeHandler({
+				poeApiKey: "key",
+				apiModelId: "openai/o3",
+				enableReasoningEffort: true,
+				reasoningEffort: "high",
+			})
+
+			const fullStream = (async function* () {
+				yield { type: "text-delta", text: "Answer" }
+			})()
+
+			mockStreamText.mockReturnValue({
+				fullStream,
+				usage: Promise.resolve({ inputTokens: 10, outputTokens: 5 }),
+			})
+
+			for await (const _ of handler.createMessage("system", [{ role: "user" as const, content: "reason" }])) {
+				/* drain */
+			}
+
+			expect(mockStreamText).toHaveBeenCalledWith(
+				expect.objectContaining({
+					providerOptions: {
+						poe: {
+							reasoningEffort: "high",
+							reasoningSummary: "auto",
+						},
+					},
+				}),
+			)
+		})
+
+		it("does not pass providerOptions when reasoning is disabled", async () => {
+			const handler = new PoeHandler({
+				poeApiKey: "key",
+				apiModelId: "anthropic/claude-sonnet-4",
+				enableReasoningEffort: false,
+			})
+
+			const fullStream = (async function* () {
+				yield { type: "text-delta", text: "Answer" }
+			})()
+
+			mockStreamText.mockReturnValue({
+				fullStream,
+				usage: Promise.resolve({ inputTokens: 1, outputTokens: 1 }),
+			})
+
+			for await (const _ of handler.createMessage("system", [{ role: "user" as const, content: "hi" }])) {
+				/* drain */
+			}
+
+			const callArgs = mockStreamText.mock.calls[0][0]
+			expect(callArgs.providerOptions).toBeUndefined()
+			expect(callArgs.temperature).toBeUndefined()
+		})
+
+		it("uses default thinking budget when not specified", async () => {
+			const handler = new PoeHandler({
+				poeApiKey: "key",
+				apiModelId: "anthropic/claude-sonnet-4",
+				enableReasoningEffort: true,
+			})
+
+			const fullStream = (async function* () {})()
+			mockStreamText.mockReturnValue({
+				fullStream,
+				usage: Promise.resolve({ inputTokens: 0, outputTokens: 0 }),
+			})
+
+			for await (const _ of handler.createMessage("system", [{ role: "user" as const, content: "hi" }])) {
+				/* drain */
+			}
+
+			const callArgs = mockStreamText.mock.calls[0][0]
+			expect(callArgs.providerOptions).toEqual({
+				poe: {
+					reasoningBudgetTokens: expect.any(Number),
+				},
+			})
+			expect(callArgs.providerOptions.poe.reasoningBudgetTokens + callArgs.maxOutputTokens).toBe(
+				handler.getModel().info.maxTokens,
+			)
+		})
+	})
+
+	describe("completePrompt", () => {
+		it("returns generated text", async () => {
+			const handler = new PoeHandler({ poeApiKey: "key", apiModelId: "openai/gpt-4o" })
+
+			mockGenerateText.mockResolvedValue({ text: "generated response" })
+
+			const result = await handler.completePrompt("complete this")
+
+			expect(result).toBe("generated response")
+			expect(mockGenerateText).toHaveBeenCalledWith(
+				expect.objectContaining({
+					model: mockLanguageModel,
+					prompt: "complete this",
+				}),
+			)
+		})
+	})
+})
--- a/src/api/providers/tests/xai.spec.ts
+++ b/src/api/providers/tests/xai.spec.ts
@ -10,14 +10,16 @@ vitest.mock("@roo-code/telemetry", () => ({
 	},
 }))

-const mockCreate = vitest.fn()
+const mockResponsesCreate = vitest.fn()

 vitest.mock("openai", () => {
 	const mockConstructor = vitest.fn()

 	return {
 		__esModule: true,
-		default: mockConstructor.mockImplementation(() => ({ chat: { completions: { create: mockCreate } } })),
+		default: mockConstructor.mockImplementation(() => ({
+			responses: { create: mockResponsesCreate },
+		})),
 	}
 })

@ -28,16 +30,30 @@ import { xaiDefaultModelId, xaiModels } from "@roo-code/types"

 import { XAIHandler } from "../xai"

+// Helper to create an async iterable from events
+function mockStream(events: any[]) {
+	return {
+		[Symbol.asyncIterator]: () => {
+			let index = 0
+			return {
+				async next() {
+					if (index < events.length) {
+						return { done: false, value: events[index++] }
+					}
+					return { done: true, value: undefined }
+				},
+			}
+		},
+	}
+}
+
 describe("XAIHandler", () => {
 	let handler: XAIHandler

 	beforeEach(() => {
-		// Reset all mocks
 		vi.clearAllMocks()
-		mockCreate.mockClear()
+		mockResponsesCreate.mockClear()
 		mockCaptureException.mockClear()
-
-		// Create handler with mock
 		handler = new XAIHandler({})
 	})

@ -50,14 +66,9 @@ describe("XAIHandler", () => {
 	})

 	it("should use the provided API key", () => {
-		// Clear mocks before this specific test
 		vi.clearAllMocks()
-
-		// Create a handler with our API key
 		const xaiApiKey = "test-api-key"
 		new XAIHandler({ xaiApiKey })
-
-		// Verify the OpenAI constructor was called with our API key
 		expect(OpenAI).toHaveBeenCalledWith(
 			expect.objectContaining({
 				apiKey: xaiApiKey,
@ -71,111 +82,41 @@ describe("XAIHandler", () => {
 		expect(model.info).toEqual(xaiModels[xaiDefaultModelId])
 	})

-	test("should return specified model when valid model is provided", () => {
+	it("should return specified model when valid model is provided", () => {
 		const testModelId = "grok-3"
 		const handlerWithModel = new XAIHandler({ apiModelId: testModelId })
 		const model = handlerWithModel.getModel()
-
 		expect(model.id).toBe(testModelId)
 		expect(model.info).toEqual(xaiModels[testModelId])
 	})

-	it("should include reasoning_effort parameter for mini models", async () => {
-		const miniModelHandler = new XAIHandler({
-			apiModelId: "grok-3-mini",
-			reasoningEffort: "high",
-		})
+	it("should use Responses API (client.responses.create)", async () => {
+		mockResponsesCreate.mockResolvedValueOnce(mockStream([]))

-		// Setup mock for streaming response
-		mockCreate.mockImplementationOnce(() => {
-			return {
-				[Symbol.asyncIterator]: () => ({
-					async next() {
-						return { done: true }
-					},
-				}),
-			}
-		})
+		const stream = handler.createMessage("test prompt", [])
+		await stream.next()

-		// Start generating a message
-		const messageGenerator = miniModelHandler.createMessage("test prompt", [])
-		await messageGenerator.next() // Start the generator
-
-		// Check that reasoning_effort was included
-		expect(mockCreate).toHaveBeenCalledWith(
+		expect(mockResponsesCreate).toHaveBeenCalledWith(
 			expect.objectContaining({
-				reasoning_effort: "high",
+				model: xaiDefaultModelId,
+				instructions: "test prompt",
+				stream: true,
+				store: false,
+				include: ["reasoning.encrypted_content"],
 			}),
 		)
 	})

-	it("should not include reasoning_effort parameter for non-mini models", async () => {
-		const regularModelHandler = new XAIHandler({
-			apiModelId: "grok-3",
-			reasoningEffort: "high",
-		})
-
-		// Setup mock for streaming response
-		mockCreate.mockImplementationOnce(() => {
-			return {
-				[Symbol.asyncIterator]: () => ({
-					async next() {
-						return { done: true }
-					},
-				}),
-			}
-		})
-
-		// Start generating a message
-		const messageGenerator = regularModelHandler.createMessage("test prompt", [])
-		await messageGenerator.next() // Start the generator
-
-		// Check call args for reasoning_effort
-		const calls = mockCreate.mock.calls
-		const lastCall = calls[calls.length - 1][0]
-		expect(lastCall).not.toHaveProperty("reasoning_effort")
-	})
-
-	it("completePrompt method should return text from OpenAI API", async () => {
-		const expectedResponse = "This is a test response"
-		mockCreate.mockResolvedValueOnce({ choices: [{ message: { content: expectedResponse } }] })
-
-		const result = await handler.completePrompt("test prompt")
-		expect(result).toBe(expectedResponse)
-	})
-
-	it("should handle errors in completePrompt", async () => {
-		const errorMessage = "API error"
-		mockCreate.mockRejectedValueOnce(new Error(errorMessage))
-
-		await expect(handler.completePrompt("test prompt")).rejects.toThrow(`xAI completion error: ${errorMessage}`)
-	})
-
 	it("createMessage should yield text content from stream", async () => {
 		const testContent = "This is test content"

-		// Setup mock for streaming response
-		mockCreate.mockImplementationOnce(() => {
-			return {
-				[Symbol.asyncIterator]: () => ({
-					next: vi
-						.fn()
-						.mockResolvedValueOnce({
-							done: false,
-							value: {
-								choices: [{ delta: { content: testContent } }],
-							},
-						})
-						.mockResolvedValueOnce({ done: true }),
-				}),
-			}
-		})
+		mockResponsesCreate.mockResolvedValueOnce(
+			mockStream([{ type: "response.output_text.delta", delta: testContent }]),
+		)

-		// Create and consume the stream
 		const stream = handler.createMessage("system prompt", [])
 		const firstChunk = await stream.next()

-		// Verify the content
 		expect(firstChunk.done).toBe(false)
 		expect(firstChunk.value).toEqual({
 			type: "text",
@ -186,28 +127,13 @@ describe("XAIHandler", () => {
 	it("createMessage should yield reasoning content from stream", async () => {
 		const testReasoning = "Test reasoning content"

-		// Setup mock for streaming response
-		mockCreate.mockImplementationOnce(() => {
-			return {
-				[Symbol.asyncIterator]: () => ({
-					next: vi
-						.fn()
-						.mockResolvedValueOnce({
-							done: false,
-							value: {
-								choices: [{ delta: { reasoning_content: testReasoning } }],
-							},
-						})
-						.mockResolvedValueOnce({ done: true }),
-				}),
-			}
-		})
+		mockResponsesCreate.mockResolvedValueOnce(
+			mockStream([{ type: "response.reasoning_text.delta", delta: testReasoning }]),
+		)

-		// Create and consume the stream
 		const stream = handler.createMessage("system prompt", [])
 		const firstChunk = await stream.next()

-		// Verify the reasoning content
 		expect(firstChunk.done).toBe(false)
 		expect(firstChunk.value).toEqual({
 			type: "reasoning",
@ -215,373 +141,158 @@ describe("XAIHandler", () => {
 		})
 	})

-	it("createMessage should yield usage data from stream", async () => {
-		// Setup mock for streaming response that includes usage data
-		mockCreate.mockImplementationOnce(() => {
-			return {
-				[Symbol.asyncIterator]: () => ({
-					next: vi
-						.fn()
-						.mockResolvedValueOnce({
-							done: false,
-							value: {
-								choices: [{ delta: {} }], // Needs to have choices array to avoid error
-								usage: {
-									prompt_tokens: 10,
-									completion_tokens: 20,
-									cache_read_input_tokens: 5,
-									cache_creation_input_tokens: 15,
-								},
-							},
-						})
-						.mockResolvedValueOnce({ done: true }),
-				}),
-			}
-		})
+	it("createMessage should yield usage data from response.completed", async () => {
+		mockResponsesCreate.mockResolvedValueOnce(
+			mockStream([
+				{
+					type: "response.completed",
+					response: {
+						usage: {
+							input_tokens: 10,
+							output_tokens: 20,
+							input_tokens_details: { cached_tokens: 5 },
+							output_tokens_details: { reasoning_tokens: 8 },
+						},
+					},
+				},
+			]),
+		)

-		// Create and consume the stream
 		const stream = handler.createMessage("system prompt", [])
 		const firstChunk = await stream.next()

-		// Verify the usage data
 		expect(firstChunk.done).toBe(false)
-		expect(firstChunk.value).toEqual({
-			type: "usage",
-			inputTokens: 10,
-			outputTokens: 20,
-			cacheReadTokens: 5,
-			cacheWriteTokens: 15,
-		})
-	})
-
-	it("createMessage should pass correct parameters to OpenAI client", async () => {
-		// Setup a handler with specific model
-		const modelId = "grok-3"
-		const modelInfo = xaiModels[modelId]
-		const handlerWithModel = new XAIHandler({ apiModelId: modelId })
-
-		// Setup mock for streaming response
-		mockCreate.mockImplementationOnce(() => {
-			return {
-				[Symbol.asyncIterator]: () => ({
-					async next() {
-						return { done: true }
-					},
-				}),
-			}
-		})
-
-		// System prompt and messages
-		const systemPrompt = "Test system prompt"
-		const messages: Anthropic.Messages.MessageParam[] = [{ role: "user", content: "Test message" }]
-
-		// Start generating a message
-		const messageGenerator = handlerWithModel.createMessage(systemPrompt, messages)
-		await messageGenerator.next() // Start the generator
-
-		// Check that all parameters were passed correctly
-		expect(mockCreate).toHaveBeenCalledWith(
+		expect(firstChunk.value).toEqual(
 			expect.objectContaining({
-				model: modelId,
-				max_tokens: modelInfo.maxTokens,
-				temperature: 0,
-				messages: expect.arrayContaining([{ role: "system", content: systemPrompt }]),
-				stream: true,
-				stream_options: { include_usage: true },
+				type: "usage",
+				inputTokens: 10,
+				outputTokens: 20,
+				cacheReadTokens: 5,
+				reasoningTokens: 8,
 			}),
 		)
 	})

-	describe("Native Tool Calling", () => {
+	it("createMessage should yield tool_call from output_item.done", async () => {
+		mockResponsesCreate.mockResolvedValueOnce(
+			mockStream([
+				{
+					type: "response.output_item.done",
+					item: {
+						type: "function_call",
+						call_id: "call_123",
+						name: "test_tool",
+						arguments: '{"arg1":"value"}',
+					},
+				},
+			]),
+		)
+
+		const stream = handler.createMessage("system prompt", [])
+		const firstChunk = await stream.next()
+
+		expect(firstChunk.done).toBe(false)
+		expect(firstChunk.value).toEqual({
+			type: "tool_call",
+			id: "call_123",
+			name: "test_tool",
+			arguments: '{"arg1":"value"}',
+		})
+	})
+
+	it("should include tools in Responses API format", async () => {
 		const testTools = [
 			{
 				type: "function" as const,
 				function: {
 					name: "test_tool",
 					description: "A test tool",
-					parameters: {
-						type: "object",
-						properties: {
-							arg1: { type: "string", description: "First argument" },
-						},
-						required: ["arg1"],
-					},
+					parameters: { type: "object", properties: { arg1: { type: "string" } }, required: ["arg1"] },
 				},
 			},
 		]

-		it("should include tools in request when model supports native tools and tools are provided (native is default)", async () => {
-			const handlerWithTools = new XAIHandler({ apiModelId: "grok-3" })
+		mockResponsesCreate.mockResolvedValueOnce(mockStream([]))

-			mockCreate.mockImplementationOnce(() => {
-				return {
-					[Symbol.asyncIterator]: () => ({
-						async next() {
-							return { done: true }
-						},
-					}),
-				}
-			})
-
-			const messageGenerator = handlerWithTools.createMessage("test prompt", [], {
-				taskId: "test-task-id",
-				tools: testTools,
-			})
-			await messageGenerator.next()
-
-			expect(mockCreate).toHaveBeenCalledWith(
-				expect.objectContaining({
-					tools: expect.arrayContaining([
-						expect.objectContaining({
-							type: "function",
-							function: expect.objectContaining({
-								name: "test_tool",
-							}),
-						}),
-					]),
-					parallel_tool_calls: true,
-				}),
-			)
+		const stream = handler.createMessage("test prompt", [], {
+			taskId: "test-task-id",
+			tools: testTools,
 		})
+		await stream.next()

-		it("should include tool_choice when provided", async () => {
-			const handlerWithTools = new XAIHandler({ apiModelId: "grok-3" })
-
-			mockCreate.mockImplementationOnce(() => {
-				return {
-					[Symbol.asyncIterator]: () => ({
-						async next() {
-							return { done: true }
-						},
+		expect(mockResponsesCreate).toHaveBeenCalledWith(
+			expect.objectContaining({
+				tools: [
+					expect.objectContaining({
+						type: "function",
+						name: "test_tool",
+						description: "A test tool",
+						strict: true,
 					}),
-				}
-			})
-
-			const messageGenerator = handlerWithTools.createMessage("test prompt", [], {
-				taskId: "test-task-id",
-				tools: testTools,
+				],
 				tool_choice: "auto",
-			})
-			await messageGenerator.next()
+				parallel_tool_calls: true,
+			}),
+		)
+	})

-			expect(mockCreate).toHaveBeenCalledWith(
-				expect.objectContaining({
-					tool_choice: "auto",
+	it("completePrompt should return text from Responses API", async () => {
+		const expectedResponse = "This is a test response"
+		mockResponsesCreate.mockResolvedValueOnce({
+			output_text: expectedResponse,
+		})
+
+		const result = await handler.completePrompt("test prompt")
+		expect(result).toBe(expectedResponse)
+	})
+
+	it("should handle errors in completePrompt", async () => {
+		const errorMessage = "API error"
+		mockResponsesCreate.mockRejectedValueOnce(new Error(errorMessage))
+
+		await expect(handler.completePrompt("test prompt")).rejects.toThrow(`xAI completion error: ${errorMessage}`)
+	})
+
+	it("should include reasoning_effort for mini models", async () => {
+		const miniModelHandler = new XAIHandler({
+			apiModelId: "grok-3-mini",
+			reasoningEffort: "high",
+		})
+
+		mockResponsesCreate.mockResolvedValueOnce(mockStream([]))
+
+		const stream = miniModelHandler.createMessage("test prompt", [])
+		await stream.next()
+
+		expect(mockResponsesCreate).toHaveBeenCalledWith(
+			expect.objectContaining({
+				reasoning: expect.objectContaining({
+					reasoning_effort: "high",
 				}),
-			)
+			}),
+		)
+	})
+
+	it("should not include reasoning for non-mini models", async () => {
+		const regularHandler = new XAIHandler({
+			apiModelId: "grok-3",
+			reasoningEffort: "high",
 		})

-		it("should always include tools and tool_choice (tools are guaranteed to be present after ALWAYS_AVAILABLE_TOOLS)", async () => {
-			const handlerWithTools = new XAIHandler({ apiModelId: "grok-3" })
+		mockResponsesCreate.mockResolvedValueOnce(mockStream([]))

-			mockCreate.mockImplementationOnce(() => {
-				return {
-					[Symbol.asyncIterator]: () => ({
-						async next() {
-							return { done: true }
-						},
-					}),
-				}
-			})
+		const stream = regularHandler.createMessage("test prompt", [])
+		await stream.next()

-			const messageGenerator = handlerWithTools.createMessage("test prompt", [], {
-				taskId: "test-task-id",
-			})
-			await messageGenerator.next()
+		const callArgs = mockResponsesCreate.mock.calls[mockResponsesCreate.mock.calls.length - 1][0]
+		expect(callArgs).not.toHaveProperty("reasoning")
+	})

-			// Tools are now always present (minimum 6 from ALWAYS_AVAILABLE_TOOLS)
-			const callArgs = mockCreate.mock.calls[mockCreate.mock.calls.length - 1][0]
-			expect(callArgs).toHaveProperty("tools")
-			expect(callArgs).toHaveProperty("tool_choice")
-			expect(callArgs).toHaveProperty("parallel_tool_calls", true)
-		})
+	it("should handle errors in createMessage", async () => {
+		const errorMessage = "Stream error"
+		mockResponsesCreate.mockRejectedValueOnce(new Error(errorMessage))

-		it("should yield tool_call_partial chunks during streaming", async () => {
-			const handlerWithTools = new XAIHandler({ apiModelId: "grok-3" })
-
-			mockCreate.mockImplementationOnce(() => {
-				return {
-					[Symbol.asyncIterator]: () => ({
-						next: vi
-							.fn()
-							.mockResolvedValueOnce({
-								done: false,
-								value: {
-									choices: [
-										{
-											delta: {
-												tool_calls: [
-													{
-														index: 0,
-														id: "call_123",
-														function: {
-															name: "test_tool",
-															arguments: '{"arg1":',
-														},
-													},
-												],
-											},
-										},
-									],
-								},
-							})
-							.mockResolvedValueOnce({
-								done: false,
-								value: {
-									choices: [
-										{
-											delta: {
-												tool_calls: [
-													{
-														index: 0,
-														function: {
-															arguments: '"value"}',
-														},
-													},
-												],
-											},
-										},
-									],
-								},
-							})
-							.mockResolvedValueOnce({ done: true }),
-					}),
-				}
-			})
-
-			const stream = handlerWithTools.createMessage("test prompt", [], {
-				taskId: "test-task-id",
-				tools: testTools,
-			})
-
-			const chunks = []
-			for await (const chunk of stream) {
-				chunks.push(chunk)
-			}
-
-			expect(chunks).toContainEqual({
-				type: "tool_call_partial",
-				index: 0,
-				id: "call_123",
-				name: "test_tool",
-				arguments: '{"arg1":',
-			})
-
-			expect(chunks).toContainEqual({
-				type: "tool_call_partial",
-				index: 0,
-				id: undefined,
-				name: undefined,
-				arguments: '"value"}',
-			})
-		})
-
-		it("should set parallel_tool_calls based on metadata", async () => {
-			const handlerWithTools = new XAIHandler({ apiModelId: "grok-3" })
-
-			mockCreate.mockImplementationOnce(() => {
-				return {
-					[Symbol.asyncIterator]: () => ({
-						async next() {
-							return { done: true }
-						},
-					}),
-				}
-			})
-
-			const messageGenerator = handlerWithTools.createMessage("test prompt", [], {
-				taskId: "test-task-id",
-				tools: testTools,
-				parallelToolCalls: true,
-			})
-			await messageGenerator.next()
-
-			expect(mockCreate).toHaveBeenCalledWith(
-				expect.objectContaining({
-					parallel_tool_calls: true,
-				}),
-			)
-		})
-
-		it("should yield tool_call_end events when finish_reason is tool_calls", async () => {
-			// Import NativeToolCallParser to set up state
-			const { NativeToolCallParser } = await import("../../../core/assistant-message/NativeToolCallParser")
-
-			// Clear any previous state
-			NativeToolCallParser.clearRawChunkState()
-
-			const handlerWithTools = new XAIHandler({ apiModelId: "grok-3" })
-
-			mockCreate.mockImplementationOnce(() => {
-				return {
-					[Symbol.asyncIterator]: () => ({
-						next: vi
-							.fn()
-							.mockResolvedValueOnce({
-								done: false,
-								value: {
-									choices: [
-										{
-											delta: {
-												tool_calls: [
-													{
-														index: 0,
-														id: "call_xai_test",
-														function: {
-															name: "test_tool",
-															arguments: '{"arg1":"value"}',
-														},
-													},
-												],
-											},
-										},
-									],
-								},
-							})
-							.mockResolvedValueOnce({
-								done: false,
-								value: {
-									choices: [
-										{
-											delta: {},
-											finish_reason: "tool_calls",
-										},
-									],
-									usage: { prompt_tokens: 10, completion_tokens: 5, total_tokens: 15 },
-								},
-							})
-							.mockResolvedValueOnce({ done: true }),
-					}),
-				}
-			})
-
-			const stream = handlerWithTools.createMessage("test prompt", [], {
-				taskId: "test-task-id",
-				tools: testTools,
-			})
-
-			const chunks = []
-			for await (const chunk of stream) {
-				// Simulate what Task.ts does: when we receive tool_call_partial,
-				// process it through NativeToolCallParser to populate rawChunkTracker
-				if (chunk.type === "tool_call_partial") {
-					NativeToolCallParser.processRawChunk({
-						index: chunk.index,
-						id: chunk.id,
-						name: chunk.name,
-						arguments: chunk.arguments,
-					})
-				}
-				chunks.push(chunk)
-			}
-
-			// Should have tool_call_partial and tool_call_end
-			const partialChunks = chunks.filter((chunk) => chunk.type === "tool_call_partial")
-			const endChunks = chunks.filter((chunk) => chunk.type === "tool_call_end")
-
-			expect(partialChunks).toHaveLength(1)
-			expect(endChunks).toHaveLength(1)
-			expect(endChunks[0].id).toBe("call_xai_test")
-		})
+		const stream = handler.createMessage("test prompt", [])
+		await expect(stream.next()).rejects.toThrow(`xAI completion error: ${errorMessage}`)
 	})
 })
--- a/src/api/providers/fetchers/tests/poe.spec.ts
+++ b/src/api/providers/fetchers/tests/poe.spec.ts
@ -0,0 +1,131 @@
+import { getPoeModels } from "../poe"
+
+const mockFetchPoeModels = vitest.fn()
+const mockGetModels = vitest.fn()
+
+vitest.mock("ai-sdk-provider-poe/code", () => ({
+	fetchPoeModels: (...args: unknown[]) => mockFetchPoeModels(...args),
+	getModels: () => mockGetModels(),
+}))
+
+describe("getPoeModels", () => {
+	beforeEach(() => vitest.clearAllMocks())
+
+	it("converts PoeModelInfo to ModelRecord", async () => {
+		mockFetchPoeModels.mockResolvedValue([])
+		mockGetModels.mockReturnValue([
+			{
+				id: "anthropic/claude-sonnet-4",
+				rawId: "claude-sonnet-4",
+				contextWindow: 200_000,
+				maxOutputTokens: 8192,
+				supportsImages: true,
+				supportsPromptCache: true,
+				supportsReasoningBudget: true,
+				pricing: {
+					inputPerMillion: 3,
+					outputPerMillion: 15,
+					cacheReadPerMillion: 0.3,
+					cacheWritePerMillion: 3.75,
+				},
+			},
+			{
+				id: "openai/gpt-4o",
+				rawId: "gpt-4o",
+				contextWindow: 128_000,
+				maxOutputTokens: 16_384,
+				supportsImages: true,
+				supportsPromptCache: false,
+				pricing: {
+					inputPerMillion: 2.5,
+					outputPerMillion: 10,
+				},
+			},
+		])
+
+		const models = await getPoeModels("test-key")
+
+		expect(mockFetchPoeModels).toHaveBeenCalledWith({ apiKey: "test-key", baseURL: undefined })
+
+		expect(models["anthropic/claude-sonnet-4"]).toEqual({
+			contextWindow: 200_000,
+			maxTokens: 8192,
+			supportsImages: true,
+			supportsPromptCache: true,
+			supportsReasoningBudget: true,
+			inputPrice: 3,
+			outputPrice: 15,
+			cacheReadsPrice: 0.3,
+			cacheWritesPrice: 3.75,
+		})
+
+		expect(models["openai/gpt-4o"]).toEqual({
+			contextWindow: 128_000,
+			maxTokens: 16_384,
+			supportsImages: true,
+			supportsPromptCache: false,
+			inputPrice: 2.5,
+			outputPrice: 10,
+		})
+	})
+
+	it("passes baseURL when provided", async () => {
+		mockFetchPoeModels.mockResolvedValue([])
+		mockGetModels.mockReturnValue([])
+
+		await getPoeModels("key", "https://custom.api.com/v1")
+
+		expect(mockFetchPoeModels).toHaveBeenCalledWith({ apiKey: "key", baseURL: "https://custom.api.com/v1" })
+	})
+
+	it("returns empty record on empty response", async () => {
+		mockFetchPoeModels.mockResolvedValue([])
+		mockGetModels.mockReturnValue([])
+
+		const models = await getPoeModels("key")
+
+		expect(models).toEqual({})
+	})
+
+	it("handles models with missing optional fields", async () => {
+		mockFetchPoeModels.mockResolvedValue([])
+		mockGetModels.mockReturnValue([
+			{
+				id: "some-model",
+				rawId: "some-model",
+				contextWindow: 4096,
+				maxOutputTokens: 1024,
+				supportsImages: false,
+				supportsPromptCache: false,
+			},
+		])
+
+		const models = await getPoeModels("key")
+
+		expect(models["some-model"]).toEqual({
+			contextWindow: 4096,
+			maxTokens: 1024,
+			supportsImages: false,
+			supportsPromptCache: false,
+		})
+	})
+
+	it("maps supportsReasoningEffort when present", async () => {
+		mockFetchPoeModels.mockResolvedValue([])
+		mockGetModels.mockReturnValue([
+			{
+				id: "openai/o3",
+				rawId: "o3",
+				contextWindow: 200_000,
+				maxOutputTokens: 100_000,
+				supportsImages: true,
+				supportsPromptCache: false,
+				supportsReasoningEffort: ["low", "medium", "high"],
+			},
+		])
+
+		const models = await getPoeModels("key")
+
+		expect(models["openai/o3"].supportsReasoningEffort).toEqual(["low", "medium", "high"])
+	})
+})
--- a/src/api/providers/fetchers/modelCache.ts
+++ b/src/api/providers/fetchers/modelCache.ts
@ -24,6 +24,7 @@ import { getLiteLLMModels } from "./litellm"
 import { GetModelsOptions } from "../../../shared/api"
 import { getOllamaModels } from "./ollama"
 import { getLMStudioModels } from "./lmstudio"
+import { getPoeModels } from "./poe"
 import { getRooModels } from "./roo"

 const memoryCache = new NodeCache({ stdTTL: 5 * 60, checkperiod: 5 * 60 })
@ -91,6 +92,9 @@ async function fetchModelsFromProvider(options: GetModelsOptions): Promise<Model
 			models = await getRooModels(rooBaseUrl, options.apiKey)
 			break
 		}
+		case "poe":
+			models = await getPoeModels(options.apiKey, options.baseUrl)
+			break
 		default: {
 			// Ensures router is exhaustively checked if RouterName is a strict union.
 			const exhaustiveCheck: never = provider
--- a/src/api/providers/fetchers/poe.ts
+++ b/src/api/providers/fetchers/poe.ts
@ -0,0 +1,44 @@
+import type { ModelInfo, ModelRecord } from "@roo-code/types"
+import { fetchPoeModels, getModels } from "ai-sdk-provider-poe/code"
+
+export async function getPoeModels(apiKey?: string, baseURL?: string): Promise<ModelRecord> {
+	try {
+		// fetchPoeModels populates the internal model store, then getModels()
+		// returns only code-capable models with camelCase fields.
+		await fetchPoeModels({ apiKey, baseURL })
+		const poeModels = getModels()
+		const models: ModelRecord = {}
+
+		for (const m of poeModels) {
+			// The library's applyReasoningFallbacks workaround sets
+			// supportsReasoningEffort to boolean `true` for any model that
+			// supports /v1/responses, even when the model has no actual
+			// reasoning capability (e.g. Haiku 3/3.5). Only trust the value
+			// when it is an explicit array of effort levels.
+			const effort = Array.isArray(m.supportsReasoningEffort) ? m.supportsReasoningEffort : undefined
+			const info: ModelInfo = {
+				contextWindow: m.contextWindow,
+				maxTokens: m.maxOutputTokens,
+				supportsImages: m.supportsImages,
+				supportsPromptCache: m.supportsPromptCache,
+				...(m.supportsReasoningBudget && { supportsReasoningBudget: m.supportsReasoningBudget }),
+				...(effort && {
+					supportsReasoningEffort: effort as ModelInfo["supportsReasoningEffort"],
+				}),
+				...(m.pricing?.inputPerMillion != null && { inputPrice: m.pricing.inputPerMillion }),
+				...(m.pricing?.outputPerMillion != null && { outputPrice: m.pricing.outputPerMillion }),
+				...(m.pricing?.cacheReadPerMillion != null && { cacheReadsPrice: m.pricing.cacheReadPerMillion }),
+				...(m.pricing?.cacheWritePerMillion != null && { cacheWritesPrice: m.pricing.cacheWritePerMillion }),
+			}
+
+			models[m.id] = info
+		}
+
+		return models
+	} catch (error) {
+		console.error(
+			`[Poe] Error fetching models: ${JSON.stringify(error, Object.getOwnPropertyNames(error as object), 2)}`,
+		)
+		return {}
+	}
+}
--- a/src/api/providers/index.ts
+++ b/src/api/providers/index.ts
@ -14,6 +14,7 @@ export { OpenAiHandler } from "./openai"
 export { OpenAICompatibleHandler } from "./openai-compatible"
 export type { OpenAICompatibleConfig } from "./openai-compatible"
 export { OpenRouterHandler } from "./openrouter"
+export { PoeHandler } from "./poe"
 export { QwenCodeHandler } from "./qwen-code"
 export { RequestyHandler } from "./requesty"
 export { SambaNovaHandler } from "./sambanova"
--- a/src/api/providers/poe.ts
+++ b/src/api/providers/poe.ts
@ -0,0 +1,151 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+import { createPoe, type PoeProvider, type PoeScopedProviderOptions } from "ai-sdk-provider-poe"
+import { extractUsageMetrics, mapToolChoice } from "ai-sdk-provider-poe/code"
+import { streamText, generateText, type ToolSet } from "ai"
+
+import {
+	poeDefaultModelId,
+	getPoeDefaultModelInfo,
+	type ModelInfo,
+	type ReasoningEffortExtended,
+	ApiProviderError,
+} from "@roo-code/types"
+import { TelemetryService } from "@roo-code/telemetry"
+
+import { shouldUseReasoningBudget, shouldUseReasoningEffort, type ApiHandlerOptions } from "../../shared/api"
+
+import { convertToAiSdkMessages, convertToolsForAiSdk, processAiSdkStreamPart } from "../transform/ai-sdk"
+import { ApiStream } from "../transform/stream"
+
+import { BaseProvider } from "./base-provider"
+import type { SingleCompletionHandler, ApiHandlerCreateMessageMetadata } from "../index"
+import { getModelsFromCache } from "./fetchers/modelCache"
+
+const DEFAULT_THINKING_BUDGET = 8192
+
+export class PoeHandler extends BaseProvider implements SingleCompletionHandler {
+	protected options: ApiHandlerOptions
+	private poe: PoeProvider
+
+	constructor(options: ApiHandlerOptions) {
+		super()
+		this.options = options
+		this.poe = createPoe({
+			apiKey: options.poeApiKey ?? "not-provided",
+			baseURL: options.poeBaseUrl || undefined,
+		})
+	}
+
+	override getModel() {
+		const id = this.options.apiModelId ?? poeDefaultModelId
+		const cached = getModelsFromCache("poe")
+		const info: ModelInfo = cached?.[id] ?? getPoeDefaultModelInfo()
+		return { id, info }
+	}
+
+	override async *createMessage(
+		systemPrompt: string,
+		messages: Anthropic.Messages.MessageParam[],
+		metadata?: ApiHandlerCreateMessageMetadata,
+	): ApiStream {
+		const { id, info } = this.getModel()
+		const languageModel = this.poe(id)
+
+		const aiSdkMessages = convertToAiSdkMessages(messages)
+		const openAiTools = this.convertToolsForOpenAI(metadata?.tools)
+		const aiSdkTools = convertToolsForAiSdk(openAiTools) as ToolSet | undefined
+
+		const useBudget = shouldUseReasoningBudget({ model: info, settings: this.options })
+		const useEffort = !useBudget && shouldUseReasoningEffort({ model: info, settings: this.options })
+
+		// Only pass temperature when the user explicitly configured it.
+		let temperature: number | undefined = this.options.modelTemperature ?? undefined
+		let maxOutputTokens: number | undefined
+		const providerOptions: NonNullable<Parameters<typeof streamText>[0]["providerOptions"]> & {
+			poe?: PoeScopedProviderOptions
+		} = {}
+
+		if (useBudget) {
+			const requestedBudget = this.options.modelMaxThinkingTokens ?? DEFAULT_THINKING_BUDGET
+			// maxOutputTokens is the text-only budget; reasoningBudgetTokens is
+			// separate, so total output = maxOutputTokens + reasoningBudgetTokens.
+			maxOutputTokens = this.options.modelMaxTokens ?? Math.max(0, (info.maxTokens ?? 0) - requestedBudget)
+			providerOptions.poe = {
+				reasoningBudgetTokens: requestedBudget,
+			}
+			temperature = 1.0
+		} else if (useEffort) {
+			let effort = (this.options.reasoningEffort ?? info.reasoningEffort ?? "medium") as ReasoningEffortExtended
+			// Validate that the effort level is actually supported by the current model
+			const supportedEfforts = info.supportsReasoningEffort
+			if (Array.isArray(supportedEfforts) && !supportedEfforts.includes(effort as any)) {
+				effort = (info.reasoningEffort as ReasoningEffortExtended) ?? "medium"
+			}
+			providerOptions.poe = {
+				reasoningEffort: effort,
+				reasoningSummary: "auto",
+			}
+			if (this.options.modelMaxTokens) {
+				maxOutputTokens = this.options.modelMaxTokens
+			}
+		}
+
+		let result
+		try {
+			result = streamText({
+				model: languageModel,
+				system: systemPrompt,
+				messages: aiSdkMessages,
+				temperature,
+				maxOutputTokens,
+				tools: aiSdkTools,
+				toolChoice: mapToolChoice(metadata?.tool_choice as any),
+				...(Object.keys(providerOptions).length > 0 && { providerOptions }),
+			})
+		} catch (error) {
+			const errorMessage = error instanceof Error ? error.message : String(error)
+			TelemetryService.instance.captureException(new ApiProviderError(errorMessage, "poe", id, "createMessage"))
+			throw new Error(`Poe completion error: ${errorMessage}`)
+		}
+
+		try {
+			for await (const part of result.fullStream) {
+				for (const chunk of processAiSdkStreamPart(part)) {
+					yield chunk
+				}
+			}
+
+			const usage = await result.usage
+			if (usage) {
+				const metrics = extractUsageMetrics(usage as any)
+				yield {
+					type: "usage" as const,
+					inputTokens: metrics.inputTokens,
+					outputTokens: metrics.outputTokens,
+					cacheReadTokens: metrics.cacheReadTokens,
+					cacheWriteTokens: metrics.cacheWriteTokens,
+					reasoningTokens: metrics.reasoningTokens,
+				}
+			}
+		} catch (error) {
+			const errorMessage = error instanceof Error ? error.message : String(error)
+			TelemetryService.instance.captureException(new ApiProviderError(errorMessage, "poe", id, "createMessage"))
+			throw new Error(`Poe streaming error: ${errorMessage}`)
+		}
+	}
+
+	async completePrompt(prompt: string): Promise<string> {
+		const { id } = this.getModel()
+		try {
+			const { text } = await generateText({
+				model: this.poe(id),
+				prompt,
+			})
+			return text
+		} catch (error) {
+			const errorMessage = error instanceof Error ? error.message : String(error)
+			TelemetryService.instance.captureException(new ApiProviderError(errorMessage, "poe", id, "completePrompt"))
+			throw new Error(`Poe completion error: ${errorMessage}`)
+		}
+	}
+}
--- a/src/api/providers/qwen-code.ts
+++ b/src/api/providers/qwen-code.ts
@ -69,6 +69,12 @@ export class QwenCodeHandler extends BaseProvider implements SingleCompletionHan
 			this.client = new OpenAI({
 				apiKey: "dummy-key-will-be-replaced",
 				baseURL: "https://dashscope.aliyuncs.com/compatible-mode/v1",
+				defaultHeaders: {
+					"User-Agent": `QwenCode/1.0.0 (${os.platform()}; ${os.arch()})`,
+					"X-DashScope-CacheControl": "enable",
+					"X-DashScope-UserAgent": `QwenCode/1.0.0 (${os.platform()}; ${os.arch()})`,
+					"X-DashScope-AuthType": "qwen-oauth",
+				},
 			})
 		}
 		return this.client
--- a/src/api/providers/xai.ts
+++ b/src/api/providers/xai.ts
@ -4,17 +4,18 @@ import OpenAI from "openai"
 import { type XAIModelId, xaiDefaultModelId, xaiModels, ApiProviderError } from "@roo-code/types"
 import { TelemetryService } from "@roo-code/telemetry"

-import { NativeToolCallParser } from "../../core/assistant-message/NativeToolCallParser"
 import type { ApiHandlerOptions } from "../../shared/api"

 import { ApiStream } from "../transform/stream"
-import { convertToOpenAiMessages } from "../transform/openai-format"
+import { convertToResponsesApiInput } from "../transform/responses-api-input"
+import { processResponsesApiStream, createUsageNormalizer } from "../transform/responses-api-stream"
 import { getModelParams } from "../transform/model-params"

 import { DEFAULT_HEADERS } from "./constants"
 import { BaseProvider } from "./base-provider"
 import type { SingleCompletionHandler, ApiHandlerCreateMessageMetadata } from "../index"
 import { handleOpenAIError } from "./utils/openai-error-handler"
+import { isMcpTool } from "../../utils/mcp-name"

 const XAI_DEFAULT_TEMPERATURE = 0

@ -53,118 +54,108 @@ export class XAIHandler extends BaseProvider implements SingleCompletionHandler
 		return { id, info, ...params }
 	}

+	/**
+	 * Convert tools from OpenAI Chat Completions format to Responses API format.
+	 * Chat Completions: { type: "function", function: { name, description, parameters } }
+	 * Responses API: { type: "function", name, description, parameters }
+	 *
+	 * Uses base provider's convertToolSchemaForOpenAI() for schema hardening
+	 * (additionalProperties: false, ensureAllRequired) and handles MCP tools.
+	 */
+	private mapResponseTools(tools?: any[]): any[] | undefined {
+		const converted = this.convertToolsForOpenAI(tools)
+		if (!converted?.length) {
+			return undefined
+		}
+		return converted
+			.filter((tool) => tool?.type === "function")
+			.map((tool) => {
+				const isMcp = isMcpTool(tool.function.name)
+				return {
+					type: "function",
+					name: tool.function.name,
+					description: tool.function.description,
+					parameters: isMcp
+						? tool.function.parameters
+						: this.convertToolSchemaForOpenAI(tool.function.parameters),
+					strict: !isMcp,
+				}
+			})
+	}
+
 	override async *createMessage(
 		systemPrompt: string,
 		messages: Anthropic.Messages.MessageParam[],
 		metadata?: ApiHandlerCreateMessageMetadata,
 	): ApiStream {
-		const { id: modelId, info: modelInfo, reasoning } = this.getModel()
+		const model = this.getModel()

-		// Use the OpenAI-compatible API.
-		const requestOptions = {
-			model: modelId,
-			max_tokens: modelInfo.maxTokens,
-			temperature: this.options.modelTemperature ?? XAI_DEFAULT_TEMPERATURE,
-			messages: [
-				{ role: "system", content: systemPrompt },
-				...convertToOpenAiMessages(messages),
-			] as OpenAI.Chat.ChatCompletionMessageParam[],
-			stream: true as const,
-			stream_options: { include_usage: true },
-			...(reasoning && reasoning),
-			tools: this.convertToolsForOpenAI(metadata?.tools),
-			tool_choice: metadata?.tool_choice,
-			parallel_tool_calls: metadata?.parallelToolCalls ?? true,
+		// Convert directly from Anthropic format to Responses API input format
+		const input = convertToResponsesApiInput(messages)
+		const responseTools = this.mapResponseTools(metadata?.tools)
+
+		// Build request options
+		const requestBody: Record<string, any> = {
+			model: model.id,
+			instructions: systemPrompt,
+			input: input,
+			stream: true,
+			store: false, // Don't store responses server-side for privacy
+			include: ["reasoning.encrypted_content"],
 		}

-		let stream
+		if (model.maxTokens) {
+			requestBody.max_output_tokens = model.maxTokens
+		}
+
+		if (model.temperature !== undefined) {
+			requestBody.temperature = model.temperature
+		}
+
+		if (responseTools) {
+			requestBody.tools = responseTools
+			// Cast tool_choice since metadata uses Chat Completions types but Responses API has its own type
+			requestBody.tool_choice = (metadata?.tool_choice ?? "auto") as any
+			requestBody.parallel_tool_calls = metadata?.parallelToolCalls ?? true
+		}
+
+		// Pass reasoning effort for models that support it (e.g., mini models)
+		if (model.reasoning) {
+			requestBody.reasoning = model.reasoning
+		}
+
+		let stream: AsyncIterable<any>
 		try {
-			stream = await this.client.chat.completions.create(requestOptions)
+			stream = (await this.client.responses.create({
+				...requestBody,
+				stream: true,
+			} as any)) as unknown as AsyncIterable<any>
 		} catch (error) {
 			const errorMessage = error instanceof Error ? error.message : String(error)
-			const apiError = new ApiProviderError(errorMessage, this.providerName, modelId, "createMessage")
+			const apiError = new ApiProviderError(errorMessage, this.providerName, model.id, "createMessage")
 			TelemetryService.instance.captureException(apiError)
 			throw handleOpenAIError(error, this.providerName)
 		}

-		for await (const chunk of stream) {
-			const delta = chunk.choices[0]?.delta
-			const finishReason = chunk.choices[0]?.finish_reason
-
-			if (delta?.content) {
-				yield {
-					type: "text",
-					text: delta.content,
-				}
-			}
-
-			if (delta && "reasoning_content" in delta && delta.reasoning_content) {
-				yield {
-					type: "reasoning",
-					text: delta.reasoning_content as string,
-				}
-			}
-
-			// Handle tool calls in stream - emit partial chunks for NativeToolCallParser
-			if (delta?.tool_calls) {
-				for (const toolCall of delta.tool_calls) {
-					yield {
-						type: "tool_call_partial",
-						index: toolCall.index,
-						id: toolCall.id,
-						name: toolCall.function?.name,
-						arguments: toolCall.function?.arguments,
-					}
-				}
-			}
-
-			// Process finish_reason to emit tool_call_end events
-			// This ensures tool calls are finalized even if the stream doesn't properly close
-			if (finishReason) {
-				const endEvents = NativeToolCallParser.processFinishReason(finishReason)
-				for (const event of endEvents) {
-					yield event
-				}
-			}
-
-			if (chunk.usage) {
-				// Extract detailed token information if available
-				// First check for prompt_tokens_details structure (real API response)
-				const promptDetails = "prompt_tokens_details" in chunk.usage ? chunk.usage.prompt_tokens_details : null
-				const cachedTokens = promptDetails && "cached_tokens" in promptDetails ? promptDetails.cached_tokens : 0
-
-				// Fall back to direct fields in usage (used in test mocks)
-				const readTokens =
-					cachedTokens ||
-					("cache_read_input_tokens" in chunk.usage ? (chunk.usage as any).cache_read_input_tokens : 0)
-				const writeTokens =
-					"cache_creation_input_tokens" in chunk.usage ? (chunk.usage as any).cache_creation_input_tokens : 0
-
-				yield {
-					type: "usage",
-					inputTokens: chunk.usage.prompt_tokens || 0,
-					outputTokens: chunk.usage.completion_tokens || 0,
-					cacheReadTokens: readTokens,
-					cacheWriteTokens: writeTokens,
-				}
-			}
-		}
+		const normalizeUsage = createUsageNormalizer()
+		yield* processResponsesApiStream(stream, normalizeUsage)
 	}

 	async completePrompt(prompt: string): Promise<string> {
-		const { id: modelId, reasoning } = this.getModel()
+		const model = this.getModel()

 		try {
-			const response = await this.client.chat.completions.create({
-				model: modelId,
-				messages: [{ role: "user", content: prompt }],
-				...(reasoning && reasoning),
+			const response = await this.client.responses.create({
+				model: model.id,
+				input: [{ role: "user", content: [{ type: "input_text", text: prompt }] }],
+				store: false,
 			})

-			return response.choices[0]?.message.content || ""
+			// output_text is a convenience field on the Responses API response
+			return response.output_text || ""
 		} catch (error) {
 			const errorMessage = error instanceof Error ? error.message : String(error)
-			const apiError = new ApiProviderError(errorMessage, this.providerName, modelId, "completePrompt")
+			const apiError = new ApiProviderError(errorMessage, this.providerName, model.id, "completePrompt")
 			TelemetryService.instance.captureException(apiError)
 			throw handleOpenAIError(error, this.providerName)
 		}
--- a/src/api/transform/tests/responses-api-input.spec.ts
+++ b/src/api/transform/tests/responses-api-input.spec.ts
@ -0,0 +1,338 @@
+import type { Anthropic } from "@anthropic-ai/sdk"
+import { convertToResponsesApiInput } from "../responses-api-input"
+
+describe("convertToResponsesApiInput", () => {
+	it("should return empty array for empty messages", () => {
+		expect(convertToResponsesApiInput([])).toEqual([])
+	})
+
+	describe("string content messages", () => {
+		it("should convert string content to input_text", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [{ role: "user", content: "Hello" }]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toEqual([{ role: "user", content: [{ type: "input_text", text: "Hello" }] }])
+		})
+
+		it("should convert assistant string content to output_text message format", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [{ role: "assistant", content: "Hi there" }]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toEqual([
+				{
+					type: "message",
+					role: "assistant",
+					content: [{ type: "output_text", text: "Hi there" }],
+				},
+			])
+		})
+	})
+
+	describe("user messages with content blocks", () => {
+		it("should convert text blocks to input_text", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "user",
+					content: [{ type: "text", text: "What is this?" }],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toEqual([{ role: "user", content: [{ type: "input_text", text: "What is this?" }] }])
+		})
+
+		it("should convert image blocks to input_image", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "user",
+					content: [
+						{
+							type: "image",
+							source: { type: "base64", media_type: "image/png", data: "abc123" },
+						},
+					],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toEqual([
+				{
+					role: "user",
+					content: [
+						{
+							type: "input_image",
+							detail: "auto",
+							image_url: "data:image/png;base64,abc123",
+						},
+					],
+				},
+			])
+		})
+
+		it("should convert tool_result to function_call_output", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "user",
+					content: [
+						{
+							type: "tool_result",
+							tool_use_id: "tool_123",
+							content: "Result text",
+						},
+					],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toEqual([
+				{
+					type: "function_call_output",
+					call_id: "tool_123",
+					output: "Result text",
+				},
+			])
+		})
+
+		it("should use (empty) for empty tool_result content", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "user",
+					content: [
+						{
+							type: "tool_result",
+							tool_use_id: "tool_123",
+							content: "",
+						},
+					],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toEqual([
+				{
+					type: "function_call_output",
+					call_id: "tool_123",
+					output: "(empty)",
+				},
+			])
+		})
+
+		it("should extract text from array tool_result content", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "user",
+					content: [
+						{
+							type: "tool_result",
+							tool_use_id: "tool_123",
+							content: [
+								{ type: "text", text: "Line 1" },
+								{ type: "text", text: "Line 2" },
+							],
+						},
+					],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toEqual([
+				{
+					type: "function_call_output",
+					call_id: "tool_123",
+					output: "Line 1\nLine 2",
+				},
+			])
+		})
+
+		it("should flush pending user content before tool_result", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "user",
+					content: [
+						{ type: "text", text: "Here is context" },
+						{
+							type: "tool_result",
+							tool_use_id: "tool_123",
+							content: "Done",
+						},
+					],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toEqual([
+				{ role: "user", content: [{ type: "input_text", text: "Here is context" }] },
+				{ type: "function_call_output", call_id: "tool_123", output: "Done" },
+			])
+		})
+	})
+
+	describe("assistant messages with content blocks", () => {
+		it("should convert text blocks to output_text messages", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "assistant",
+					content: [{ type: "text", text: "Here is my response" }],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toEqual([
+				{
+					type: "message",
+					role: "assistant",
+					content: [{ type: "output_text", text: "Here is my response" }],
+				},
+			])
+		})
+
+		it("should convert tool_use to function_call", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "tool_use",
+							id: "call_abc",
+							name: "read_file",
+							input: { path: "/tmp/test.txt" },
+						},
+					],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toEqual([
+				{
+					type: "function_call",
+					call_id: "call_abc",
+					name: "read_file",
+					arguments: '{"path":"/tmp/test.txt"}',
+				},
+			])
+		})
+
+		it("should handle tool_use with string input", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "tool_use",
+							id: "call_abc",
+							name: "run_command",
+							input: '{"cmd":"ls"}' as any,
+						},
+					],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result[0]).toEqual(
+				expect.objectContaining({
+					type: "function_call",
+					arguments: '{"cmd":"ls"}',
+				}),
+			)
+		})
+
+		it("should handle mixed text and tool_use in assistant message", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{
+					role: "assistant",
+					content: [
+						{ type: "text", text: "Let me read that file" },
+						{
+							type: "tool_use",
+							id: "call_abc",
+							name: "read_file",
+							input: { path: "/tmp/test.txt" },
+						},
+					],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toHaveLength(2)
+			expect(result[0]).toEqual(
+				expect.objectContaining({
+					type: "message",
+					role: "assistant",
+					content: [{ type: "output_text", text: "Let me read that file" }],
+				}),
+			)
+			expect(result[1]).toEqual(
+				expect.objectContaining({
+					type: "function_call",
+					name: "read_file",
+				}),
+			)
+		})
+	})
+
+	describe("multi-turn conversations", () => {
+		it("should handle a complete tool use cycle", () => {
+			const messages: Anthropic.Messages.MessageParam[] = [
+				{ role: "user", content: "Read /tmp/test.txt" },
+				{
+					role: "assistant",
+					content: [
+						{
+							type: "tool_use",
+							id: "call_1",
+							name: "read_file",
+							input: { path: "/tmp/test.txt" },
+						},
+					],
+				},
+				{
+					role: "user",
+					content: [
+						{
+							type: "tool_result",
+							tool_use_id: "call_1",
+							content: "file contents here",
+						},
+					],
+				},
+				{
+					role: "assistant",
+					content: [{ type: "text", text: "The file contains: file contents here" }],
+				},
+			]
+
+			const result = convertToResponsesApiInput(messages)
+
+			expect(result).toHaveLength(4)
+			expect(result[0]).toEqual({ role: "user", content: [{ type: "input_text", text: "Read /tmp/test.txt" }] })
+			expect(result[1]).toEqual(
+				expect.objectContaining({ type: "function_call", call_id: "call_1", name: "read_file" }),
+			)
+			expect(result[2]).toEqual(
+				expect.objectContaining({
+					type: "function_call_output",
+					call_id: "call_1",
+					output: "file contents here",
+				}),
+			)
+			expect(result[3]).toEqual(
+				expect.objectContaining({
+					type: "message",
+					role: "assistant",
+				}),
+			)
+		})
+	})
+})
--- a/src/api/transform/tests/responses-api-stream.spec.ts
+++ b/src/api/transform/tests/responses-api-stream.spec.ts
@ -0,0 +1,392 @@
+import { processResponsesApiStream, createUsageNormalizer } from "../responses-api-stream"
+
+// Helper to create an async iterable from events
+async function* mockStream(events: any[]) {
+	for (const event of events) {
+		yield event
+	}
+}
+
+// Helper to collect all chunks from a stream
+async function collectChunks(stream: AsyncGenerator<any>) {
+	const chunks = []
+	for await (const chunk of stream) {
+		chunks.push(chunk)
+	}
+	return chunks
+}
+
+const noopUsage = () => undefined
+
+describe("processResponsesApiStream", () => {
+	describe("text deltas", () => {
+		it("should yield text chunk for response.output_text.delta", async () => {
+			const stream = mockStream([{ type: "response.output_text.delta", delta: "Hello world" }])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([{ type: "text", text: "Hello world" }])
+		})
+
+		it("should yield text chunk for response.text.delta", async () => {
+			const stream = mockStream([{ type: "response.text.delta", delta: "Hello" }])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([{ type: "text", text: "Hello" }])
+		})
+
+		it("should skip text delta with empty delta", async () => {
+			const stream = mockStream([{ type: "response.output_text.delta", delta: "" }])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([])
+		})
+	})
+
+	describe("reasoning deltas", () => {
+		it("should yield reasoning chunk for response.reasoning_text.delta", async () => {
+			const stream = mockStream([{ type: "response.reasoning_text.delta", delta: "Let me think..." }])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([{ type: "reasoning", text: "Let me think..." }])
+		})
+
+		it("should yield reasoning chunk for response.reasoning.delta", async () => {
+			const stream = mockStream([{ type: "response.reasoning.delta", delta: "Step 1" }])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([{ type: "reasoning", text: "Step 1" }])
+		})
+
+		it("should yield reasoning chunk for response.reasoning_summary_text.delta", async () => {
+			const stream = mockStream([{ type: "response.reasoning_summary_text.delta", delta: "Summary" }])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([{ type: "reasoning", text: "Summary" }])
+		})
+
+		it("should yield reasoning chunk for response.reasoning_summary.delta", async () => {
+			const stream = mockStream([{ type: "response.reasoning_summary.delta", delta: "Summary" }])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([{ type: "reasoning", text: "Summary" }])
+		})
+	})
+
+	describe("tool calls", () => {
+		it("should yield tool_call for function_call in output_item.done", async () => {
+			const stream = mockStream([
+				{
+					type: "response.output_item.done",
+					item: {
+						type: "function_call",
+						call_id: "call_123",
+						name: "read_file",
+						arguments: '{"path":"/tmp/test.txt"}',
+					},
+				},
+			])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([
+				{
+					type: "tool_call",
+					id: "call_123",
+					name: "read_file",
+					arguments: '{"path":"/tmp/test.txt"}',
+				},
+			])
+		})
+
+		it("should yield tool_call for tool_call type in output_item.done", async () => {
+			const stream = mockStream([
+				{
+					type: "response.output_item.done",
+					item: {
+						type: "tool_call",
+						tool_call_id: "call_456",
+						name: "write_file",
+						arguments: '{"path":"/tmp/out.txt"}',
+					},
+				},
+			])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([
+				{
+					type: "tool_call",
+					id: "call_456",
+					name: "write_file",
+					arguments: '{"path":"/tmp/out.txt"}',
+				},
+			])
+		})
+
+		it("should handle object arguments by JSON.stringifying", async () => {
+			const stream = mockStream([
+				{
+					type: "response.output_item.done",
+					item: {
+						type: "function_call",
+						call_id: "call_789",
+						name: "test",
+						input: { key: "value" },
+					},
+				},
+			])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks[0].arguments).toBe('{"key":"value"}')
+		})
+
+		it("should skip tool_call with missing call_id or name", async () => {
+			const stream = mockStream([
+				{
+					type: "response.output_item.done",
+					item: { type: "function_call", call_id: "", name: "test", arguments: "{}" },
+				},
+				{
+					type: "response.output_item.done",
+					item: { type: "function_call", call_id: "call_1", name: "", arguments: "{}" },
+				},
+			])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([])
+		})
+
+		it("should yield tool_call_partial for function_call_arguments.delta", async () => {
+			const stream = mockStream([
+				{
+					type: "response.function_call_arguments.delta",
+					call_id: "call_123",
+					name: "read_file",
+					delta: '{"path":',
+					index: 0,
+				},
+			])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([
+				{
+					type: "tool_call_partial",
+					index: 0,
+					id: "call_123",
+					name: "read_file",
+					arguments: '{"path":',
+				},
+			])
+		})
+	})
+
+	describe("completion and usage", () => {
+		it("should yield usage from response.completed", async () => {
+			const mockNormalize = (usage: any) => ({
+				type: "usage" as const,
+				inputTokens: usage.input_tokens,
+				outputTokens: usage.output_tokens,
+			})
+
+			const stream = mockStream([
+				{
+					type: "response.completed",
+					response: { usage: { input_tokens: 100, output_tokens: 50 } },
+				},
+			])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, mockNormalize))
+
+			expect(chunks).toEqual([{ type: "usage", inputTokens: 100, outputTokens: 50 }])
+		})
+
+		it("should yield usage from response.done", async () => {
+			const mockNormalize = (usage: any) => ({
+				type: "usage" as const,
+				inputTokens: usage.input_tokens,
+				outputTokens: usage.output_tokens,
+			})
+
+			const stream = mockStream([
+				{
+					type: "response.done",
+					response: { usage: { input_tokens: 200, output_tokens: 100 } },
+				},
+			])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, mockNormalize))
+
+			expect(chunks).toEqual([{ type: "usage", inputTokens: 200, outputTokens: 100 }])
+		})
+
+		it("should not yield usage when normalizer returns undefined", async () => {
+			const stream = mockStream([
+				{
+					type: "response.completed",
+					response: { usage: null },
+				},
+			])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([])
+		})
+	})
+
+	describe("unknown events", () => {
+		it("should silently ignore unknown event types", async () => {
+			const stream = mockStream([
+				{ type: "response.created" },
+				{ type: "response.in_progress" },
+				{ type: "response.output_item.added", item: { type: "message" } },
+				{ type: "response.content_part.added" },
+			])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, noopUsage))
+
+			expect(chunks).toEqual([])
+		})
+	})
+
+	describe("full conversation stream", () => {
+		it("should handle a complete stream with reasoning, text, and usage", async () => {
+			const mockNormalize = (usage: any) => ({
+				type: "usage" as const,
+				inputTokens: usage.input_tokens,
+				outputTokens: usage.output_tokens,
+			})
+
+			const stream = mockStream([
+				{ type: "response.reasoning_text.delta", delta: "Thinking..." },
+				{ type: "response.reasoning_text.delta", delta: " done." },
+				{ type: "response.output_text.delta", delta: "The answer is " },
+				{ type: "response.output_text.delta", delta: "42." },
+				{
+					type: "response.completed",
+					response: { usage: { input_tokens: 50, output_tokens: 30 } },
+				},
+			])
+
+			const chunks = await collectChunks(processResponsesApiStream(stream, mockNormalize))
+
+			expect(chunks).toEqual([
+				{ type: "reasoning", text: "Thinking..." },
+				{ type: "reasoning", text: " done." },
+				{ type: "text", text: "The answer is " },
+				{ type: "text", text: "42." },
+				{ type: "usage", inputTokens: 50, outputTokens: 30 },
+			])
+		})
+	})
+})
+
+describe("createUsageNormalizer", () => {
+	it("should return undefined for null/undefined usage", () => {
+		const normalize = createUsageNormalizer()
+		expect(normalize(null)).toBeUndefined()
+		expect(normalize(undefined)).toBeUndefined()
+	})
+
+	it("should extract input and output tokens", () => {
+		const normalize = createUsageNormalizer()
+
+		const result = normalize({ input_tokens: 100, output_tokens: 50 })
+
+		expect(result).toEqual(
+			expect.objectContaining({
+				type: "usage",
+				inputTokens: 100,
+				outputTokens: 50,
+			}),
+		)
+	})
+
+	it("should extract cached tokens from input_tokens_details", () => {
+		const normalize = createUsageNormalizer()
+
+		const result = normalize({
+			input_tokens: 100,
+			output_tokens: 50,
+			input_tokens_details: { cached_tokens: 30 },
+		})
+
+		expect(result?.cacheReadTokens).toBe(30)
+	})
+
+	it("should extract cache write tokens", () => {
+		const normalize = createUsageNormalizer()
+
+		const result = normalize({
+			input_tokens: 100,
+			output_tokens: 50,
+			cache_creation_input_tokens: 15,
+		})
+
+		expect(result?.cacheWriteTokens).toBe(15)
+	})
+
+	it("should extract reasoning tokens from output_tokens_details", () => {
+		const normalize = createUsageNormalizer()
+
+		const result = normalize({
+			input_tokens: 100,
+			output_tokens: 50,
+			output_tokens_details: { reasoning_tokens: 20 },
+		})
+
+		expect(result?.reasoningTokens).toBe(20)
+	})
+
+	it("should not include reasoningTokens when not present", () => {
+		const normalize = createUsageNormalizer()
+
+		const result = normalize({ input_tokens: 100, output_tokens: 50 })
+
+		expect(result).not.toHaveProperty("reasoningTokens")
+	})
+
+	it("should compute totalCost when calculateCost is provided", () => {
+		const calculateCost = (input: number, output: number, cached: number) => 0.42
+		const normalize = createUsageNormalizer(calculateCost)
+
+		const result = normalize({ input_tokens: 100, output_tokens: 50 })
+
+		expect(result?.totalCost).toBe(0.42)
+	})
+
+	it("should not include totalCost when calculateCost is not provided", () => {
+		const normalize = createUsageNormalizer()
+
+		const result = normalize({ input_tokens: 100, output_tokens: 50 })
+
+		expect(result).not.toHaveProperty("totalCost")
+	})
+
+	it("should handle Chat Completions style field names as fallback", () => {
+		const normalize = createUsageNormalizer()
+
+		const result = normalize({
+			prompt_tokens: 100,
+			completion_tokens: 50,
+			prompt_tokens_details: { cached_tokens: 10 },
+		})
+
+		expect(result).toEqual(
+			expect.objectContaining({
+				inputTokens: 100,
+				outputTokens: 50,
+				cacheReadTokens: 10,
+			}),
+		)
+	})
+})
--- a/src/api/transform/responses-api-input.ts
+++ b/src/api/transform/responses-api-input.ts
@ -0,0 +1,118 @@
+import { Anthropic } from "@anthropic-ai/sdk"
+
+/**
+ * Converts Anthropic-format messages to the OpenAI Responses API input format.
+ *
+ * Key differences from Chat Completions format:
+ * - Content parts use { type: "input_text" } instead of { type: "text" }
+ * - Images use { type: "input_image" } instead of { type: "image_url" }
+ * - Tool results use { type: "function_call_output", call_id } instead of { role: "tool", tool_call_id }
+ * - Tool uses become { type: "function_call", call_id, name, arguments } items
+ * - System prompt goes via the `instructions` parameter, not as a message
+ *
+ * @param messages - Array of Anthropic MessageParam objects
+ * @returns Array of Responses API input items
+ */
+export function convertToResponsesApiInput(messages: Anthropic.Messages.MessageParam[]): any[] {
+	const input: any[] = []
+
+	for (const message of messages) {
+		if (typeof message.content === "string") {
+			if (message.role === "assistant") {
+				// Assistant messages use output_text in the Responses API format
+				input.push({
+					type: "message",
+					role: "assistant",
+					content: [{ type: "output_text", text: message.content }],
+				})
+			} else {
+				input.push({
+					role: message.role,
+					content: [{ type: "input_text", text: message.content }],
+				})
+			}
+			continue
+		}
+
+		if (message.role === "assistant") {
+			for (const part of message.content) {
+				switch (part.type) {
+					case "text":
+						input.push({
+							type: "message",
+							role: "assistant",
+							content: [{ type: "output_text", text: part.text }],
+						})
+						break
+					case "tool_use":
+						input.push({
+							type: "function_call",
+							call_id: part.id,
+							name: part.name,
+							arguments: typeof part.input === "string" ? part.input : JSON.stringify(part.input ?? {}),
+						})
+						break
+					case "thinking":
+						// Include reasoning if it has content
+						if ((part as any).thinking && (part as any).thinking.trim().length > 0) {
+							input.push({
+								type: "message",
+								role: "assistant",
+								content: [{ type: "output_text", text: `[Thinking] ${(part as any).thinking}` }],
+							})
+						}
+						break
+				}
+			}
+		} else {
+			// User messages
+			const contentParts: any[] = []
+			for (const part of message.content) {
+				switch (part.type) {
+					case "text":
+						contentParts.push({ type: "input_text", text: part.text })
+						break
+					case "image":
+						contentParts.push({
+							type: "input_image",
+							detail: "auto",
+							image_url: `data:${part.source.media_type};base64,${part.source.data}`,
+						})
+						break
+					case "tool_result": {
+						// Flush any pending user content before the tool result
+						if (contentParts.length > 0) {
+							input.push({ role: "user", content: [...contentParts] })
+							contentParts.length = 0
+						}
+						// Convert tool result content
+						let output: string
+						if (typeof part.content === "string") {
+							output = part.content || "(empty)"
+						} else if (Array.isArray(part.content)) {
+							output =
+								part.content
+									.filter((c): c is Anthropic.TextBlockParam => c.type === "text")
+									.map((c) => c.text)
+									.join("\n") || "(empty)"
+						} else {
+							output = "(empty)"
+						}
+						input.push({
+							type: "function_call_output",
+							call_id: part.tool_use_id,
+							output,
+						})
+						break
+					}
+				}
+			}
+			// Flush remaining user content
+			if (contentParts.length > 0) {
+				input.push({ role: "user", content: contentParts })
+			}
+		}
+	}
+
+	return input
+}
--- a/src/api/transform/responses-api-stream.ts
+++ b/src/api/transform/responses-api-stream.ts
@ -0,0 +1,140 @@
+import type { ApiStream, ApiStreamUsageChunk } from "./stream"
+
+/**
+ * Processes Responses API stream events and yields ApiStreamChunks.
+ *
+ * This is a shared utility for providers that use OpenAI's Responses API
+ * (POST /v1/responses with stream: true). It handles the core event types:
+ *
+ * - Text deltas (response.output_text.delta)
+ * - Reasoning deltas (response.reasoning_text.delta, response.reasoning_summary_text.delta)
+ * - Tool/function calls (response.output_item.done with function_call type)
+ * - Usage data (response.completed)
+ *
+ * Provider-specific concerns (WebSocket mode, SSE fallback, duplicate detection,
+ * pending tool tracking) are intentionally left to individual providers.
+ *
+ * @param stream - AsyncIterable of Responses API stream events
+ * @param normalizeUsage - Provider-specific function to normalize usage data into ApiStreamUsageChunk
+ */
+export async function* processResponsesApiStream(
+	stream: AsyncIterable<any>,
+	normalizeUsage: (usage: any) => ApiStreamUsageChunk | undefined,
+): ApiStream {
+	for await (const event of stream) {
+		// Text content deltas
+		if (event?.type === "response.output_text.delta" || event?.type === "response.text.delta") {
+			if (event?.delta) {
+				yield { type: "text", text: event.delta }
+			}
+			continue
+		}
+
+		// Reasoning deltas
+		if (
+			event?.type === "response.reasoning_text.delta" ||
+			event?.type === "response.reasoning.delta" ||
+			event?.type === "response.reasoning_summary_text.delta" ||
+			event?.type === "response.reasoning_summary.delta"
+		) {
+			if (event?.delta) {
+				yield { type: "reasoning", text: event.delta }
+			}
+			continue
+		}
+
+		// Output item events — handle completed function calls and fallback text
+		if (event?.type === "response.output_item.done") {
+			const item = event?.item
+			if (item?.type === "function_call" || item?.type === "tool_call") {
+				const callId = item.call_id || item.tool_call_id || item.id
+				const name = item.name || item.function?.name
+				const argsRaw = item.arguments || item.function?.arguments || item.input
+				const args =
+					typeof argsRaw === "string"
+						? argsRaw
+						: argsRaw && typeof argsRaw === "object"
+							? JSON.stringify(argsRaw)
+							: ""
+
+				if (typeof callId === "string" && callId.length > 0 && typeof name === "string" && name.length > 0) {
+					yield {
+						type: "tool_call",
+						id: callId,
+						name,
+						arguments: args,
+					}
+				}
+			}
+			continue
+		}
+
+		// Function call argument deltas (for streaming tool calls)
+		if (
+			event?.type === "response.function_call_arguments.delta" ||
+			event?.type === "response.tool_call_arguments.delta"
+		) {
+			const callId = event.call_id || event.tool_call_id || event.id || event.item_id
+			const name = event.name || event.function_name
+			if (typeof callId === "string" && callId.length > 0) {
+				yield {
+					type: "tool_call_partial",
+					index: event.index ?? 0,
+					id: callId,
+					name,
+					arguments: typeof event.delta === "string" ? event.delta : "",
+				}
+			}
+			continue
+		}
+
+		// Completion events — extract usage
+		if (event?.type === "response.completed" || event?.type === "response.done") {
+			const usage = event?.response?.usage || event?.usage
+			const usageData = normalizeUsage(usage)
+			if (usageData) {
+				yield usageData
+			}
+			continue
+		}
+	}
+}
+
+/**
+ * Creates a standard usage normalizer for providers with per-token pricing.
+ * Extracts input/output tokens, cache tokens, reasoning tokens, and computes cost.
+ *
+ * @param calculateCost - Optional function to compute total cost from token counts
+ */
+export function createUsageNormalizer(
+	calculateCost?: (inputTokens: number, outputTokens: number, cacheReadTokens: number) => number,
+): (usage: any) => ApiStreamUsageChunk | undefined {
+	return (usage: any): ApiStreamUsageChunk | undefined => {
+		if (!usage) return undefined
+
+		const inputDetails = usage.input_tokens_details ?? usage.prompt_tokens_details
+		const cachedTokens = inputDetails?.cached_tokens ?? 0
+
+		const inputTokens = usage.input_tokens ?? usage.prompt_tokens ?? 0
+		const outputTokens = usage.output_tokens ?? usage.completion_tokens ?? 0
+		const cacheReadTokens = usage.cache_read_input_tokens ?? cachedTokens ?? 0
+		const cacheWriteTokens = usage.cache_creation_input_tokens ?? usage.cache_write_tokens ?? 0
+
+		const reasoningTokens =
+			typeof usage.output_tokens_details?.reasoning_tokens === "number"
+				? usage.output_tokens_details.reasoning_tokens
+				: undefined
+
+		const totalCost = calculateCost ? calculateCost(inputTokens, outputTokens, cacheReadTokens) : undefined
+
+		return {
+			type: "usage",
+			inputTokens,
+			outputTokens,
+			cacheWriteTokens,
+			cacheReadTokens,
+			...(typeof reasoningTokens === "number" ? { reasoningTokens } : {}),
+			...(typeof totalCost === "number" ? { totalCost } : {}),
+		}
+	}
+}
--- a/src/core/webview/ClineProvider.ts
+++ b/src/core/webview/ClineProvider.ts
@ -169,7 +169,7 @@ export class ClineProvider

 	public isViewLaunched = false
 	public settingsImportedAt?: number
-	public readonly latestAnnouncementId = "mar-2026-v3.51.0-gpt-54-slash-skills" // v3.51.0 OpenAI GPT-5.4 support and slash command skills
+	public readonly latestAnnouncementId = "apr-2026-v3.52.0-poe-xai-minimax" // v3.52.0 Poe provider, xAI improvements, and MiniMax fixes
 	public readonly providerSettingsManager: ProviderSettingsManager
 	public readonly customModesManager: CustomModesManager

--- a/src/core/webview/tests/ClineProvider.spec.ts
+++ b/src/core/webview/tests/ClineProvider.spec.ts
@ -2494,6 +2494,7 @@ describe("ClineProvider - Router Models", () => {
 				ollama: {},
 				lmstudio: {},
 				"vercel-ai-gateway": mockModels,
+				poe: {},
 			},
 			values: undefined,
 		})
@ -2540,6 +2541,7 @@ describe("ClineProvider - Router Models", () => {
 				lmstudio: {},
 				litellm: {},
 				"vercel-ai-gateway": mockModels,
+				poe: {},
 			},
 			values: undefined,
 		})
@ -2634,6 +2636,7 @@ describe("ClineProvider - Router Models", () => {
 				ollama: {},
 				lmstudio: {},
 				"vercel-ai-gateway": mockModels,
+				poe: {},
 			},
 			values: undefined,
 		})
--- a/src/core/webview/tests/webviewMessageHandler.spec.ts
+++ b/src/core/webview/tests/webviewMessageHandler.spec.ts
@ -346,6 +346,7 @@ describe("webviewMessageHandler - requestRouterModels", () => {
 				ollama: {},
 				lmstudio: {},
 				"vercel-ai-gateway": mockModels,
+				poe: {},
 			},
 			values: undefined,
 		})
@ -431,6 +432,7 @@ describe("webviewMessageHandler - requestRouterModels", () => {
 				ollama: {},
 				lmstudio: {},
 				"vercel-ai-gateway": mockModels,
+				poe: {},
 			},
 			values: undefined,
 		})
@ -486,6 +488,7 @@ describe("webviewMessageHandler - requestRouterModels", () => {
 				ollama: {},
 				lmstudio: {},
 				"vercel-ai-gateway": mockModels,
+				poe: {},
 			},
 			values: undefined,
 		})
--- a/src/core/webview/webviewMessageHandler.ts
+++ b/src/core/webview/webviewMessageHandler.ts
@ -955,6 +955,7 @@ export const webviewMessageHandler = async (
 						ollama: {},
 						lmstudio: {},
 						roo: {},
+						poe: {},
 					}

 			const safeGetModels = async (options: GetModelsOptions): Promise<ModelRecord> => {
@ -1018,6 +1019,21 @@ export const webviewMessageHandler = async (
 				})
 			}

+			// Poe is conditional on apiKey
+			const poeApiKey = apiConfiguration.poeApiKey || message?.values?.poeApiKey
+			const poeBaseUrl = apiConfiguration.poeBaseUrl || message?.values?.poeBaseUrl
+
+			if (poeApiKey) {
+				if (message?.values?.poeApiKey || message?.values?.poeBaseUrl) {
+					await flushModels({ provider: "poe", apiKey: poeApiKey, baseUrl: poeBaseUrl }, true)
+				}
+
+				candidates.push({
+					key: "poe",
+					options: { provider: "poe", apiKey: poeApiKey, baseUrl: poeBaseUrl },
+				})
+			}
+
 			// Apply single provider filter if specified
 			const modelFetchPromises = providerFilter
 				? candidates.filter(({ key }) => key === providerFilter)
--- a/src/package.json
+++ b/src/package.json
@ -3,7 +3,7 @@
 	"displayName": "%extension.displayName%",
 	"description": "%extension.description%",
 	"publisher": "RooVeterinaryInc",
-	"version": "3.51.1",
+	"version": "3.52.0",
 	"icon": "assets/icons/icon.png",
 	"galleryBanner": {
 		"color": "#617A91",
@ -471,6 +471,7 @@
 		"@roo-code/telemetry": "workspace:^",
 		"@roo-code/types": "workspace:^",
 		"@vscode/codicons": "^0.0.36",
+		"ai-sdk-provider-poe": "2.0.18",
 		"async-mutex": "^0.5.0",
 		"axios": "^1.12.0",
 		"cheerio": "^1.0.0",
--- a/src/shared/api.ts
+++ b/src/shared/api.ts
@ -177,6 +177,7 @@ const dynamicProviderExtras = {
 	ollama: {} as {}, // eslint-disable-line @typescript-eslint/no-empty-object-type
 	lmstudio: {} as {}, // eslint-disable-line @typescript-eslint/no-empty-object-type
 	roo: {} as { apiKey?: string; baseUrl?: string },
+	poe: {} as { apiKey?: string; baseUrl?: string },
 } as const satisfies Record<RouterName, object>

 // Build the dynamic options union from the map, intersected with CommonFetchParams
--- a/webview-ui/src/components/chat/tests/Announcement.spec.tsx
+++ b/webview-ui/src/components/chat/tests/Announcement.spec.tsx
@ -12,7 +12,7 @@ vi.mock("@src/utils/vscode", () => ({

 vi.mock("@roo/package", () => ({
 	Package: {
-		version: "3.51.0",
+		version: "3.52.0",
 	},
 }))

@ -34,9 +34,9 @@ vi.mock("@src/i18n/TranslationContext", () => ({
 			const translations: Record<string, string> = {
 				"chat:announcement.release.heading": "What's New:",
 				"chat:announcement.release.gpt54":
-					"OpenAI GPT-5.4 Support: Added OpenAI GPT-5.4 and GPT-5.3 Chat Latest so you can use the newest OpenAI chat models in Roo Code.",
+					"Poe Provider: Added Poe as an AI provider so you can access Poe models directly in Roo Code.",
 				"chat:announcement.release.slashSkills":
-					"Slash Command Skills: Skills can now be exposed as slash commands with fallback execution for faster workflows.",
+					"xAI and MiniMax Improvements: Migrated the xAI provider to the Responses API, added Grok-4.20 defaults, and fixed MiniMax model listings and context window handling for a more reliable setup.",
 			}

 			if (key === "chat:announcement.title") {
@ -49,18 +49,18 @@ vi.mock("@src/i18n/TranslationContext", () => ({
 }))

 describe("Announcement", () => {
-	it("renders the v3.51.0 announcement title and highlights", () => {
+	it("renders the v3.52.0 announcement title and highlights", () => {
 		render(<Announcement hideAnnouncement={vi.fn()} />)

-		expect(screen.getByText("Roo Code 3.51.0 Released")).toBeInTheDocument()
+		expect(screen.getByText("Roo Code 3.52.0 Released")).toBeInTheDocument()
 		expect(
 			screen.getByText(
-				"OpenAI GPT-5.4 Support: Added OpenAI GPT-5.4 and GPT-5.3 Chat Latest so you can use the newest OpenAI chat models in Roo Code.",
+				"Poe Provider: Added Poe as an AI provider so you can access Poe models directly in Roo Code.",
 			),
 		).toBeInTheDocument()
 		expect(
 			screen.getByText(
-				"Slash Command Skills: Skills can now be exposed as slash commands with fallback execution for faster workflows.",
+				"xAI and MiniMax Improvements: Migrated the xAI provider to the Responses API, added Grok-4.20 defaults, and fixed MiniMax model listings and context window handling for a more reliable setup.",
 			),
 		).toBeInTheDocument()
 	})
--- a/webview-ui/src/components/settings/ApiOptions.tsx
+++ b/webview-ui/src/components/settings/ApiOptions.tsx
@ -1,7 +1,7 @@
 import React, { memo, useCallback, useEffect, useMemo, useState } from "react"
 import { convertHeadersToObject } from "./utils/headers"
 import { useDebounce } from "react-use"
-import { VSCodeLink } from "@vscode/webview-ui-toolkit/react"
+import { VSCodeLink, VSCodeTextField } from "@vscode/webview-ui-toolkit/react"
 import { ExternalLinkIcon } from "@radix-ui/react-icons"

 import {
@ -10,6 +10,7 @@ import {
 	isRetiredProvider,
 	DEFAULT_CONSECUTIVE_MISTAKE_LIMIT,
 	openRouterDefaultModelId,
+	poeDefaultModelId,
 	requestyDefaultModelId,
 	litellmDefaultModelId,
 	openAiNativeDefaultModelId,
@ -80,6 +81,7 @@ import {
 	OpenAICompatible,
 	OpenAICodex,
 	OpenRouter,
+	Poe,
 	QwenCode,
 	Requesty,
 	Roo,
@ -238,7 +240,7 @@ const ApiOptions = ({
 				vscode.postMessage({ type: "requestLmStudioModels" })
 			} else if (selectedProvider === "vscode-lm") {
 				vscode.postMessage({ type: "requestVsCodeLmModels" })
-			} else if (selectedProvider === "litellm" || selectedProvider === "roo") {
+			} else if (selectedProvider === "litellm" || selectedProvider === "roo" || selectedProvider === "poe") {
 				vscode.postMessage({ type: "requestRouterModels" })
 			}
 		},
@ -252,6 +254,8 @@ const ApiOptions = ({
 			apiConfiguration?.lmStudioBaseUrl,
 			apiConfiguration?.litellmBaseUrl,
 			apiConfiguration?.litellmApiKey,
+			apiConfiguration?.poeApiKey,
+			apiConfiguration?.poeBaseUrl,
 			customHeaders,
 		],
 	)
@ -356,6 +360,7 @@ const ApiOptions = ({
 							: internationalZAiDefaultModelId,
 				},
 				fireworks: { field: "apiModelId", default: fireworksDefaultModelId },
+				poe: { field: "apiModelId", default: poeDefaultModelId },
 				roo: { field: "apiModelId", default: rooDefaultModelId },
 				"vercel-ai-gateway": { field: "vercelAiGatewayModelId", default: vercelAiGatewayDefaultModelId },
 				openai: { field: "openAiModelId" },
@ -703,6 +708,16 @@ const ApiOptions = ({
 						/>
 					)}

+					{selectedProvider === "poe" && (
+						<Poe
+							apiConfiguration={apiConfiguration}
+							setApiConfigurationField={setApiConfigurationField}
+							organizationAllowList={organizationAllowList}
+							modelValidationError={modelValidationError}
+							simplifySettings={fromWelcomeView}
+						/>
+					)}
+
 					{selectedProvider === "roo" && (
 						<Roo
 							apiConfiguration={apiConfiguration}
@ -800,6 +815,17 @@ const ApiOptions = ({
 									}
 									onChange={(value) => setApiConfigurationField("consecutiveMistakeLimit", value)}
 								/>
+								{selectedProvider === "poe" && (
+									<VSCodeTextField
+										value={apiConfiguration?.poeBaseUrl || ""}
+										onInput={handleInputChange("poeBaseUrl")}
+										placeholder="https://api.poe.com/v1"
+										className="w-full">
+										<label className="block font-medium mb-1">
+											{t("settings:providers.poeBaseUrl")}
+										</label>
+									</VSCodeTextField>
+								)}
 								{selectedProvider === "openrouter" &&
 									openRouterModelProviders &&
 									Object.keys(openRouterModelProviders).length > 0 && (
--- a/webview-ui/src/components/settings/constants.ts
+++ b/webview-ui/src/components/settings/constants.ts
@ -65,4 +65,5 @@ export const PROVIDERS = [
 	{ value: "minimax", label: "MiniMax", proxy: false },
 	{ value: "baseten", label: "Baseten", proxy: false },
 	{ value: "unbound", label: "Unbound", proxy: false },
+	{ value: "poe", label: "Poe", proxy: false },
 ].sort((a, b) => a.label.localeCompare(b.label))
--- a/webview-ui/src/components/settings/providers/Poe.tsx
+++ b/webview-ui/src/components/settings/providers/Poe.tsx
@ -0,0 +1,166 @@
+import { useCallback, useState, useEffect, useRef } from "react"
+import { VSCodeTextField } from "@vscode/webview-ui-toolkit/react"
+import { useQueryClient } from "@tanstack/react-query"
+
+import {
+	type ProviderSettings,
+	type OrganizationAllowList,
+	type ExtensionMessage,
+	poeDefaultModelId,
+	type ProviderName,
+} from "@roo-code/types"
+
+import { RouterName } from "@roo/api"
+
+import { useAppTranslation } from "@src/i18n/TranslationContext"
+import { VSCodeButtonLink } from "@src/components/common/VSCodeButtonLink"
+import { useExtensionState } from "@src/context/ExtensionStateContext"
+import { vscode } from "@src/utils/vscode"
+import { Button } from "@src/components/ui"
+
+import { inputEventTransform } from "../transforms"
+import { ModelPicker } from "../ModelPicker"
+import { handleModelChangeSideEffects } from "../utils/providerModelConfig"
+
+type PoeProps = {
+	apiConfiguration: ProviderSettings
+	setApiConfigurationField: (field: keyof ProviderSettings, value: ProviderSettings[keyof ProviderSettings]) => void
+	organizationAllowList: OrganizationAllowList
+	modelValidationError?: string
+	simplifySettings?: boolean
+}
+
+export const Poe = ({
+	apiConfiguration,
+	setApiConfigurationField,
+	organizationAllowList,
+	modelValidationError,
+	simplifySettings,
+}: PoeProps) => {
+	const { t } = useAppTranslation()
+	const queryClient = useQueryClient()
+	const { routerModels } = useExtensionState()
+	const [refreshStatus, setRefreshStatus] = useState<"idle" | "loading" | "success" | "error">("idle")
+	const [refreshError, setRefreshError] = useState<string | undefined>()
+	const poeErrorJustReceived = useRef(false)
+
+	useEffect(() => {
+		const handleMessage = (event: MessageEvent<ExtensionMessage>) => {
+			const message = event.data
+			if (message.type === "singleRouterModelFetchResponse" && !message.success) {
+				const providerName = message.values?.provider as RouterName
+				if (providerName === "poe") {
+					poeErrorJustReceived.current = true
+					setRefreshStatus("error")
+					setRefreshError(message.error)
+				}
+			} else if (message.type === "routerModels") {
+				if (refreshStatus === "loading") {
+					if (!poeErrorJustReceived.current) {
+						setRefreshStatus("success")
+						// Invalidate the react-query router models cache so
+						// validation in ApiOptions picks up the refreshed list.
+						queryClient.invalidateQueries({ queryKey: ["routerModels"] })
+					}
+				}
+			}
+		}
+
+		window.addEventListener("message", handleMessage)
+		return () => {
+			window.removeEventListener("message", handleMessage)
+		}
+	}, [refreshStatus, queryClient])
+
+	const handleInputChange = useCallback(
+		<K extends keyof ProviderSettings, E>(
+			field: K,
+			transform: (event: E) => ProviderSettings[K] = inputEventTransform,
+		) =>
+			(event: E | Event) => {
+				setApiConfigurationField(field, transform(event as E))
+			},
+		[setApiConfigurationField],
+	)
+
+	const handleRefreshModels = useCallback(() => {
+		poeErrorJustReceived.current = false
+		setRefreshStatus("loading")
+		setRefreshError(undefined)
+
+		const key = apiConfiguration.poeApiKey
+
+		if (!key) {
+			setRefreshStatus("error")
+			setRefreshError(t("settings:providers.refreshModels.missingConfig"))
+			return
+		}
+
+		vscode.postMessage({
+			type: "requestRouterModels",
+			values: { poeApiKey: key, poeBaseUrl: apiConfiguration.poeBaseUrl },
+		})
+	}, [apiConfiguration, t])
+
+	return (
+		<>
+			<VSCodeTextField
+				value={apiConfiguration?.poeApiKey || ""}
+				type="password"
+				onInput={handleInputChange("poeApiKey")}
+				placeholder={t("settings:placeholders.apiKey")}
+				className="w-full">
+				<label className="block font-medium mb-1">{t("settings:providers.poeApiKey")}</label>
+			</VSCodeTextField>
+			<div className="text-sm text-vscode-descriptionForeground -mt-2">
+				{t("settings:providers.apiKeyStorageNotice")}
+			</div>
+			{!apiConfiguration?.poeApiKey && (
+				<VSCodeButtonLink href="https://poe.com/api_key" appearance="secondary">
+					{t("settings:providers.getPoeApiKey")}
+				</VSCodeButtonLink>
+			)}
+			<Button
+				variant="outline"
+				onClick={handleRefreshModels}
+				disabled={refreshStatus === "loading" || !apiConfiguration.poeApiKey}>
+				<div className="flex items-center gap-2">
+					{refreshStatus === "loading" ? (
+						<span className="codicon codicon-loading codicon-modifier-spin" />
+					) : (
+						<span className="codicon codicon-refresh" />
+					)}
+					{t("settings:providers.refreshModels.label")}
+				</div>
+			</Button>
+			{refreshStatus === "loading" && (
+				<div className="text-sm text-vscode-descriptionForeground">
+					{t("settings:providers.refreshModels.loading")}
+				</div>
+			)}
+			{refreshStatus === "success" && (
+				<div className="text-sm text-vscode-foreground">{t("settings:providers.refreshModels.success")}</div>
+			)}
+			{refreshStatus === "error" && (
+				<div className="text-sm text-vscode-errorForeground">
+					{refreshError || t("settings:providers.refreshModels.error")}
+				</div>
+			)}
+			<ModelPicker
+				apiConfiguration={apiConfiguration}
+				setApiConfigurationField={setApiConfigurationField}
+				defaultModelId={poeDefaultModelId}
+				models={routerModels?.poe ?? {}}
+				modelIdKey="apiModelId"
+				serviceName="Poe"
+				serviceUrl="https://poe.com"
+				organizationAllowList={organizationAllowList}
+				errorMessage={modelValidationError}
+				simplifySettings={simplifySettings}
+				onModelChange={(modelId) =>
+					handleModelChangeSideEffects("poe" as ProviderName, modelId, setApiConfigurationField)
+				}
+			/>
+		</>
+	)
+}
--- a/webview-ui/src/components/settings/providers/index.ts
+++ b/webview-ui/src/components/settings/providers/index.ts
@ -10,6 +10,7 @@ export { OpenAI } from "./OpenAI"
 export { OpenAICodex } from "./OpenAICodex"
 export { OpenAICompatible } from "./OpenAICompatible"
 export { OpenRouter } from "./OpenRouter"
+export { Poe } from "./Poe"
 export { QwenCode } from "./QwenCode"
 export { Roo } from "./Roo"
 export { Requesty } from "./Requesty"
--- a/webview-ui/src/components/settings/utils/providerModelConfig.ts
+++ b/webview-ui/src/components/settings/utils/providerModelConfig.ts
@ -150,8 +150,10 @@ export const handleModelChangeSideEffects = <K extends keyof ProviderSettings>(
 		setApiConfigurationField("awsCustomArn" as K, "" as ProviderSettings[K])
 	}

-	// All providers: Clear reasoning effort when switching models to allow
-	// the new model's default to take effect. Different models within the
-	// same provider can have different reasoning effort defaults/options.
+	// All providers: Clear reasoning settings when switching models to allow
+	// the new model's defaults to take effect. Different models within the
+	// same provider can have different reasoning defaults/options.
 	setApiConfigurationField("reasoningEffort" as K, undefined as ProviderSettings[K])
+	setApiConfigurationField("modelMaxTokens" as K, undefined as ProviderSettings[K])
+	setApiConfigurationField("modelMaxThinkingTokens" as K, undefined as ProviderSettings[K])
 }
--- a/webview-ui/src/components/ui/hooks/tests/useSelectedModel.spec.ts
+++ b/webview-ui/src/components/ui/hooks/tests/useSelectedModel.spec.ts
@ -11,6 +11,8 @@ import {
 	BEDROCK_1M_CONTEXT_MODEL_IDS,
 	litellmDefaultModelInfo,
 	openAiModelInfoSaneDefaults,
+	minimaxDefaultModelId,
+	minimaxModels,
 } from "@roo-code/types"

 import { useSelectedModel } from "../useSelectedModel"
@ -725,4 +727,51 @@ describe("useSelectedModel", () => {
 			expect(result.current.info).toEqual(customModelInfo)
 		})
 	})
+
+	describe("minimax provider", () => {
+		beforeEach(() => {
+			mockUseRouterModels.mockReturnValue({
+				data: {
+					openrouter: {},
+					requesty: {},
+					litellm: {},
+				},
+				isLoading: false,
+				isError: false,
+			} as any)
+
+			mockUseOpenRouterModelProviders.mockReturnValue({
+				data: {},
+				isLoading: false,
+				isError: false,
+			} as any)
+		})
+
+		it("should return default minimax model when no custom model is specified", () => {
+			const apiConfiguration: ProviderSettings = {
+				apiProvider: "minimax",
+			}
+
+			const wrapper = createWrapper()
+			const { result } = renderHook(() => useSelectedModel(apiConfiguration), { wrapper })
+
+			expect(result.current.provider).toBe("minimax")
+			expect(result.current.id).toBe(minimaxDefaultModelId)
+			expect(result.current.info).toEqual(minimaxModels[minimaxDefaultModelId])
+		})
+
+		it("should use custom model ID and info when model exists in minimaxModels", () => {
+			const apiConfiguration: ProviderSettings = {
+				apiProvider: "minimax",
+				apiModelId: "MiniMax-M2.7",
+			}
+
+			const wrapper = createWrapper()
+			const { result } = renderHook(() => useSelectedModel(apiConfiguration), { wrapper })
+
+			expect(result.current.provider).toBe("minimax")
+			expect(result.current.id).toBe("MiniMax-M2.7")
+			expect(result.current.info).toEqual(minimaxModels["MiniMax-M2.7"])
+		})
+	})
 })
--- a/webview-ui/src/components/ui/hooks/useSelectedModel.ts
+++ b/webview-ui/src/components/ui/hooks/useSelectedModel.ts
@ -317,6 +317,11 @@ function getSelectedModel({
 			const info = routerModels.roo?.[id]
 			return { id, info }
 		}
+		case "poe": {
+			const id = apiConfiguration.apiModelId ?? defaultModelId
+			const info = routerModels.poe?.[id]
+			return { id, info }
+		}
 		case "qwen-code": {
 			const id = apiConfiguration.apiModelId ?? defaultModelId
 			const info = qwenCodeModels[id as keyof typeof qwenCodeModels]
--- a/webview-ui/src/i18n/locales/ca/chat.json
+++ b/webview-ui/src/i18n/locales/ca/chat.json
@ -348,8 +348,8 @@
 		},
 		"release": {
 			"heading": "Què hi ha de nou:",
-			"gpt54": "Compatibilitat amb OpenAI GPT-5.4: Hem afegit OpenAI GPT-5.4 i GPT-5.3 Chat Latest perquè puguis fer servir els models de xat més nous d’OpenAI a Roo Code.",
-			"slashSkills": "Skills com a ordres slash: Ara les skills es poden exposar com a ordres slash amb execució de reserva per accelerar els fluxos de treball.",
+			"gpt54": "Proveïdor Poe: S'ha afegit Poe com a proveïdor d’IA perquè puguis accedir directament als models de Poe a Roo Code.",
+			"slashSkills": "Millores a xAI i MiniMax: S'ha migrat el proveïdor xAI a la Responses API, s'hi han afegit utilitats de transformació reutilitzables, s'han actualitzat els valors per defecte de Grok-4.20 i s'han corregit els llistats de models de MiniMax i la gestió de la finestra de context per oferir una configuració més fiable.",
 			"geminiPro": "Suport per Gemini 3.1 Pro: S'ha afegit el suport per al model Gemini 3.1 Pro i s'ha establert com a model Gemini per defecte per a un millor rendiment.",
 			"cliNdjson": "Protocol NDJSON del CLI: Nou protocol NDJSON via stdin, subcomandes de llistat i comanda d'execució modularitzada per a fluxos de treball CLI més flexibles.",
 			"cliRelease": "CLI v0.1.0: El CLI de Roo Code arriba a la seva primera versió oficial amb una interfície de comandes estable."
--- a/webview-ui/src/i18n/locales/ca/settings.json
+++ b/webview-ui/src/i18n/locales/ca/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Amplia la finestra de context a 1 milió de tokens per a Claude Sonnet 4.x / Claude Opus 4.6",
 		"basetenApiKey": "Clau API de Baseten",
 		"getBasetenApiKey": "Obtenir clau API de Baseten",
+		"poeApiKey": "Clau API de Poe",
+		"getPoeApiKey": "Obtenir clau API de Poe",
+		"poeBaseUrl": "URL base de Poe",
 		"fireworksApiKey": "Clau API de Fireworks",
 		"getFireworksApiKey": "Obtenir clau API de Fireworks",
 		"deepSeekApiKey": "Clau API de DeepSeek",
--- a/webview-ui/src/i18n/locales/de/chat.json
+++ b/webview-ui/src/i18n/locales/de/chat.json
@ -348,8 +348,8 @@
 		},
 		"release": {
 			"heading": "Was ist neu:",
-			"gpt54": "OpenAI GPT-5.4-Unterstützung: OpenAI GPT-5.4 und GPT-5.3 Chat Latest wurden hinzugefügt, damit du die neuesten OpenAI-Chatmodelle in Roo Code nutzen kannst.",
-			"slashSkills": "Slash-Command-Skills: Skills können jetzt als Slash-Commands mit Fallback-Ausführung verfügbar gemacht werden, damit Workflows schneller ablaufen.",
+			"gpt54": "Poe-Anbieter: Poe als KI-Anbieter hinzugefügt, damit du direkt in Roo Code auf Poe-Modelle zugreifen kannst.",
+			"slashSkills": "xAI- und MiniMax-Verbesserungen: Den xAI-Anbieter auf die Responses API migriert, wiederverwendbare Transformations-Utilities hinzugefügt, Grok-4.20-Standards aktualisiert und MiniMax-Modelllisten sowie die Handhabung des Kontextfensters für eine zuverlässigere Einrichtung korrigiert.",
 			"geminiPro": "Gemini 3.1 Pro Unterstützung: Gemini 3.1 Pro wurde als Modell hinzugefügt und als Standard-Gemini-Modell für bessere Leistung festgelegt.",
 			"cliNdjson": "CLI NDJSON-Protokoll: Neues NDJSON-stdin-Protokoll, List-Unterbefehle und modularisierter Run-Befehl für flexiblere CLI-Workflows.",
 			"cliRelease": "CLI v0.1.0: Das Roo Code CLI erreicht seine erste offizielle Veröffentlichung mit einer stabilen Befehlsoberfläche."
--- a/webview-ui/src/i18n/locales/de/settings.json
+++ b/webview-ui/src/i18n/locales/de/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Erweitert das Kontextfenster für Claude Sonnet 4.x / Claude Opus 4.6 auf 1 Million Token",
 		"basetenApiKey": "Baseten API-Schlüssel",
 		"getBasetenApiKey": "Baseten API-Schlüssel erhalten",
+		"poeApiKey": "Poe API-Schlüssel",
+		"getPoeApiKey": "Poe API-Schlüssel erhalten",
+		"poeBaseUrl": "Poe Basis-URL",
 		"fireworksApiKey": "Fireworks API-Schlüssel",
 		"getFireworksApiKey": "Fireworks API-Schlüssel erhalten",
 		"deepSeekApiKey": "DeepSeek API-Schlüssel",
--- a/webview-ui/src/i18n/locales/en/chat.json
+++ b/webview-ui/src/i18n/locales/en/chat.json
@ -367,8 +367,8 @@
 		},
 		"release": {
 			"heading": "What's New:",
-			"gpt54": "OpenAI GPT-5.4 Support: Added OpenAI GPT-5.4 and GPT-5.3 Chat Latest so you can use the newest OpenAI chat models in Roo Code.",
-			"slashSkills": "Slash Command Skills: Skills can now be exposed as slash commands with fallback execution for faster workflows."
+			"gpt54": "Poe Provider: Added Poe as an AI provider so you can access Poe models directly in Roo Code.",
+			"slashSkills": "xAI and MiniMax Improvements: Migrated the xAI provider to the Responses API, added Grok-4.20 defaults, and fixed MiniMax model listings and context window handling for a more reliable setup."
 		},
 		"cloudAgents": {
 			"heading": "New in the Cloud:",
--- a/webview-ui/src/i18n/locales/en/settings.json
+++ b/webview-ui/src/i18n/locales/en/settings.json
@ -443,6 +443,9 @@
 		"vertex1MContextBetaDescription": "Extends context window to 1 million tokens for Claude Sonnet 4.x / Claude Opus 4.6",
 		"basetenApiKey": "Baseten API Key",
 		"getBasetenApiKey": "Get Baseten API Key",
+		"poeApiKey": "Poe API Key",
+		"getPoeApiKey": "Get Poe API Key",
+		"poeBaseUrl": "Poe Base URL",
 		"fireworksApiKey": "Fireworks API Key",
 		"getFireworksApiKey": "Get Fireworks API Key",
 		"deepSeekApiKey": "DeepSeek API Key",
--- a/webview-ui/src/i18n/locales/es/chat.json
+++ b/webview-ui/src/i18n/locales/es/chat.json
@ -348,8 +348,8 @@
 		},
 		"release": {
 			"heading": "Qué hay de nuevo:",
-			"gpt54": "Compatibilidad con OpenAI GPT-5.4: Se añadieron OpenAI GPT-5.4 y GPT-5.3 Chat Latest para que puedas usar los modelos de chat más recientes de OpenAI en Roo Code.",
-			"slashSkills": "Skills como comandos slash: Ahora las skills pueden exponerse como comandos slash con ejecución de respaldo para activar flujos de trabajo más rápido.",
+			"gpt54": "Proveedor Poe: Se añadió Poe como proveedor de IA para que puedas acceder directamente a los modelos de Poe en Roo Code.",
+			"slashSkills": "Mejoras en xAI y MiniMax: Se migró el proveedor xAI a la Responses API, se añadieron utilidades de transformación reutilizables, se actualizaron los valores predeterminados de Grok-4.20 y se corrigieron los listados de modelos de MiniMax y el manejo de la ventana de contexto para una configuración más fiable.",
 			"geminiPro": "Soporte para Gemini 3.1 Pro: Se añadió soporte para el modelo Gemini 3.1 Pro y se estableció como modelo Gemini predeterminado para un mejor rendimiento.",
 			"cliNdjson": "Protocolo NDJSON del CLI: Nuevo protocolo NDJSON por stdin, subcomandos de listado y comando de ejecución modularizado para flujos de trabajo CLI más flexibles.",
 			"cliRelease": "CLI v0.1.0: El CLI de Roo Code alcanza su primera versión oficial con una interfaz de comandos estable."
--- a/webview-ui/src/i18n/locales/es/settings.json
+++ b/webview-ui/src/i18n/locales/es/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Amplía la ventana de contexto a 1 millón de tokens para Claude Sonnet 4.x / Claude Opus 4.6",
 		"basetenApiKey": "Clave API de Baseten",
 		"getBasetenApiKey": "Obtener clave API de Baseten",
+		"poeApiKey": "Clave API de Poe",
+		"getPoeApiKey": "Obtener clave API de Poe",
+		"poeBaseUrl": "URL base de Poe",
 		"fireworksApiKey": "Clave API de Fireworks",
 		"getFireworksApiKey": "Obtener clave API de Fireworks",
 		"deepSeekApiKey": "Clave API de DeepSeek",
--- a/webview-ui/src/i18n/locales/fr/chat.json
+++ b/webview-ui/src/i18n/locales/fr/chat.json
@ -348,8 +348,8 @@
 		},
 		"release": {
 			"heading": "Quoi de neuf :",
-			"gpt54": "Prise en charge d’OpenAI GPT-5.4 : OpenAI GPT-5.4 et GPT-5.3 Chat Latest ont été ajoutés pour que tu puisses utiliser les modèles de chat OpenAI les plus récents dans Roo Code.",
-			"slashSkills": "Skills en slash commands : Les skills peuvent maintenant être exposées comme slash commands avec une exécution de secours pour accélérer les workflows.",
+			"gpt54": "Fournisseur Poe : Poe a été ajouté comme fournisseur d’IA pour que tu puisses accéder directement aux modèles Poe dans Roo Code.",
+			"slashSkills": "Améliorations xAI et MiniMax : Le fournisseur xAI a été migré vers la Responses API, des utilitaires de transformation réutilisables ont été ajoutés, les valeurs par défaut de Grok-4.20 ont été mises à jour et les listes de modèles MiniMax ainsi que la gestion de la fenêtre de contexte ont été corrigées pour une configuration plus fiable.",
 			"geminiPro": "Support de Gemini 3.1 Pro : Le modèle Gemini 3.1 Pro a été ajouté et défini comme modèle Gemini par défaut pour de meilleures performances.",
 			"cliNdjson": "Protocole NDJSON du CLI : Nouveau protocole NDJSON via stdin, sous-commandes de liste et commande d'exécution modularisée pour des workflows CLI plus flexibles.",
 			"cliRelease": "CLI v0.1.0 : Le CLI de Roo Code atteint sa première version officielle avec une interface de commandes stable."
--- a/webview-ui/src/i18n/locales/fr/settings.json
+++ b/webview-ui/src/i18n/locales/fr/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Étend la fenêtre de contexte à 1 million de tokens pour Claude Sonnet 4.x / Claude Opus 4.6",
 		"basetenApiKey": "Clé API Baseten",
 		"getBasetenApiKey": "Obtenir la clé API Baseten",
+		"poeApiKey": "Clé API Poe",
+		"getPoeApiKey": "Obtenir la clé API Poe",
+		"poeBaseUrl": "URL de base Poe",
 		"fireworksApiKey": "Clé API Fireworks",
 		"getFireworksApiKey": "Obtenir la clé API Fireworks",
 		"deepSeekApiKey": "Clé API DeepSeek",
--- a/webview-ui/src/i18n/locales/hi/chat.json
+++ b/webview-ui/src/i18n/locales/hi/chat.json
@ -348,8 +348,8 @@
 		},
 		"release": {
 			"heading": "नया क्या है:",
-			"gpt54": "OpenAI GPT-5.4 सपोर्ट: OpenAI GPT-5.4 और GPT-5.3 Chat Latest जोड़े गए हैं, ताकि आप Roo Code में OpenAI के सबसे नए chat models इस्तेमाल कर सकें।",
-			"slashSkills": "Slash command skills: अब skills को slash commands के रूप में fallback execution के साथ expose किया जा सकता है, जिससे workflows तेज़ हो जाते हैं।",
+			"gpt54": "Poe provider: Poe को AI provider के रूप में जोड़ा गया है, ताकि आप Roo Code में सीधे Poe models तक पहुँच सकें।",
+			"slashSkills": "xAI और MiniMax में सुधार: xAI provider को Responses API पर migrate किया गया है, reusable transform utilities जोड़ी गई हैं, Grok-4.20 defaults अपडेट किए गए हैं, और ज़्यादा भरोसेमंद setup के लिए MiniMax model listings व context window handling को ठीक किया गया है।",
 			"geminiPro": "Gemini 3.1 Pro सपोर्ट: Gemini 3.1 Pro मॉडल का सपोर्ट जोड़ा गया और बेहतर प्रदर्शन के लिए इसे डिफ़ॉल्ट Gemini मॉडल के रूप में सेट किया गया।",
 			"cliNdjson": "CLI NDJSON प्रोटोकॉल: नया NDJSON stdin प्रोटोकॉल, list सबकमांड, और अधिक लचीले CLI वर्कफ़्लो के लिए मॉड्यूलर run कमांड।",
 			"cliRelease": "CLI v0.1.0: Roo Code CLI एक स्थिर कमांड इंटरफ़ेस के साथ अपनी पहली आधिकारिक रिलीज़ तक पहुँचा।"
--- a/webview-ui/src/i18n/locales/hi/settings.json
+++ b/webview-ui/src/i18n/locales/hi/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Claude Sonnet 4.x / Claude Opus 4.6 के लिए संदर्भ विंडो को 1 मिलियन टोकन तक बढ़ाता है",
 		"basetenApiKey": "Baseten API कुंजी",
 		"getBasetenApiKey": "Baseten API कुंजी प्राप्त करें",
+		"poeApiKey": "Poe API कुंजी",
+		"getPoeApiKey": "Poe API कुंजी प्राप्त करें",
+		"poeBaseUrl": "Poe बेस URL",
 		"fireworksApiKey": "Fireworks API कुंजी",
 		"getFireworksApiKey": "Fireworks API कुंजी प्राप्त करें",
 		"deepSeekApiKey": "DeepSeek API कुंजी",
--- a/webview-ui/src/i18n/locales/id/chat.json
+++ b/webview-ui/src/i18n/locales/id/chat.json
@ -377,8 +377,8 @@
 		},
 		"release": {
 			"heading": "Yang Baru:",
-			"gpt54": "Dukungan OpenAI GPT-5.4: OpenAI GPT-5.4 dan GPT-5.3 Chat Latest kini ditambahkan agar kamu bisa memakai model chat OpenAI terbaru di Roo Code.",
-			"slashSkills": "Skill sebagai slash command: Skill sekarang bisa diekspos sebagai slash command dengan eksekusi fallback supaya workflow bisa berjalan lebih cepat.",
+			"gpt54": "Provider Poe: Poe ditambahkan sebagai provider AI agar kamu bisa langsung mengakses model Poe di Roo Code.",
+			"slashSkills": "Peningkatan xAI dan MiniMax: Provider xAI dimigrasikan ke Responses API, utilitas transformasi yang bisa dipakai ulang ditambahkan, default Grok-4.20 diperbarui, dan daftar model MiniMax serta penanganan context window diperbaiki agar setup lebih andal.",
 			"geminiPro": "Dukungan Gemini 3.1 Pro: Dukungan model Gemini 3.1 Pro telah ditambahkan dan ditetapkan sebagai model Gemini default untuk performa yang lebih baik.",
 			"cliNdjson": "Protokol NDJSON CLI: Protokol NDJSON stdin baru, subperintah list, dan perintah run yang dimodularisasi untuk alur kerja CLI yang lebih fleksibel.",
 			"cliRelease": "CLI v0.1.0: CLI Roo Code mencapai rilis resmi pertamanya dengan antarmuka perintah yang stabil."
--- a/webview-ui/src/i18n/locales/id/settings.json
+++ b/webview-ui/src/i18n/locales/id/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Memperluas jendela konteks menjadi 1 juta token untuk Claude Sonnet 4.x / Claude Opus 4.6",
 		"basetenApiKey": "Baseten API Key",
 		"getBasetenApiKey": "Dapatkan Baseten API Key",
+		"poeApiKey": "Poe API Key",
+		"getPoeApiKey": "Dapatkan Poe API Key",
+		"poeBaseUrl": "Poe Base URL",
 		"fireworksApiKey": "Fireworks API Key",
 		"getFireworksApiKey": "Dapatkan Fireworks API Key",
 		"deepSeekApiKey": "DeepSeek API Key",
--- a/webview-ui/src/i18n/locales/it/chat.json
+++ b/webview-ui/src/i18n/locales/it/chat.json
@ -348,8 +348,8 @@
 		},
 		"release": {
 			"heading": "Novità:",
-			"gpt54": "Supporto per OpenAI GPT-5.4: abbiamo aggiunto OpenAI GPT-5.4 e GPT-5.3 Chat Latest così puoi usare i modelli di chat OpenAI più recenti in Roo Code.",
-			"slashSkills": "Skill come slash command: ora le skill possono essere esposte come slash command con esecuzione di fallback per attivare i workflow più rapidamente.",
+			"gpt54": "Provider Poe: abbiamo aggiunto Poe come provider AI così puoi accedere direttamente ai modelli Poe in Roo Code.",
+			"slashSkills": "Miglioramenti per xAI e MiniMax: il provider xAI è stato migrato alla Responses API, sono state aggiunte utility di trasformazione riutilizzabili, sono stati aggiornati i valori predefiniti di Grok-4.20 e sono stati corretti gli elenchi dei modelli MiniMax e la gestione della finestra di contesto per una configurazione più affidabile.",
 			"geminiPro": "Supporto Gemini 3.1 Pro: Aggiunto il supporto per il modello Gemini 3.1 Pro e impostato come modello Gemini predefinito per prestazioni migliori.",
 			"cliNdjson": "Protocollo NDJSON del CLI: Nuovo protocollo NDJSON via stdin, sottocomandi di elenco e comando di esecuzione modularizzato per flussi di lavoro CLI più flessibili.",
 			"cliRelease": "CLI v0.1.0: Il CLI di Roo Code raggiunge la sua prima release ufficiale con un'interfaccia di comandi stabile."
--- a/webview-ui/src/i18n/locales/it/settings.json
+++ b/webview-ui/src/i18n/locales/it/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Estende la finestra di contesto a 1 milione di token per Claude Sonnet 4.x / Claude Opus 4.6",
 		"basetenApiKey": "Chiave API Baseten",
 		"getBasetenApiKey": "Ottieni chiave API Baseten",
+		"poeApiKey": "Chiave API Poe",
+		"getPoeApiKey": "Ottieni chiave API Poe",
+		"poeBaseUrl": "URL base Poe",
 		"fireworksApiKey": "Chiave API Fireworks",
 		"getFireworksApiKey": "Ottieni chiave API Fireworks",
 		"deepSeekApiKey": "Chiave API DeepSeek",
--- a/webview-ui/src/i18n/locales/ja/chat.json
+++ b/webview-ui/src/i18n/locales/ja/chat.json
@ -348,8 +348,8 @@
 		},
 		"release": {
 			"heading": "新機能:",
-			"gpt54": "OpenAI GPT-5.4 サポート: OpenAI GPT-5.4 と GPT-5.3 Chat Latest を追加し、Roo Code で OpenAI の最新チャットモデルを使えるようにしました。",
-			"slashSkills": "スラッシュコマンド対応スキル: スキルをスラッシュコマンドとして公開し、フォールバック実行にも対応したことで、ワークフローをより素早く実行できます。",
+			"gpt54": "Poe プロバイダー: Poe を AI プロバイダーとして追加し、Roo Code から Poe のモデルに直接アクセスできるようにしました。",
+			"slashSkills": "xAI と MiniMax の改善: xAI プロバイダーを Responses API へ移行し、再利用可能な変換ユーティリティを追加し、Grok-4.20 のデフォルト設定を更新し、MiniMax のモデル一覧とコンテキストウィンドウの扱いを修正して、より信頼性の高いセットアップにしました。",
 			"geminiPro": "Gemini 3.1 Pro サポート: Gemini 3.1 Pro モデルのサポートを追加し、パフォーマンス向上のためデフォルトの Gemini モデルに設定しました。",
 			"cliNdjson": "CLI NDJSON プロトコル: 新しい NDJSON stdin プロトコル、list サブコマンド、より柔軟な CLI ワークフローのためのモジュール化された run コマンド。",
 			"cliRelease": "CLI v0.1.0: Roo Code CLI が安定したコマンドインターフェースで初の正式リリースに到達しました。"
--- a/webview-ui/src/i18n/locales/ja/settings.json
+++ b/webview-ui/src/i18n/locales/ja/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Claude Sonnet 4.x / Claude Opus 4.6のコンテキストウィンドウを100万トークンに拡張します",
 		"basetenApiKey": "Baseten APIキー",
 		"getBasetenApiKey": "Baseten APIキーを取得",
+		"poeApiKey": "Poe APIキー",
+		"getPoeApiKey": "Poe APIキーを取得",
+		"poeBaseUrl": "Poe ベースURL",
 		"fireworksApiKey": "Fireworks APIキー",
 		"getFireworksApiKey": "Fireworks APIキーを取得",
 		"deepSeekApiKey": "DeepSeek APIキー",
--- a/webview-ui/src/i18n/locales/ko/chat.json
+++ b/webview-ui/src/i18n/locales/ko/chat.json
@ -348,8 +348,8 @@
 		},
 		"release": {
 			"heading": "새로운 기능:",
-			"gpt54": "OpenAI GPT-5.4 지원: OpenAI GPT-5.4와 GPT-5.3 Chat Latest를 추가해 Roo Code에서 최신 OpenAI 채팅 모델을 사용할 수 있게 했어요.",
-			"slashSkills": "슬래시 명령어 스킬: 이제 스킬을 슬래시 명령어로 노출하고 fallback 실행도 지원해 워크플로를 더 빠르게 실행할 수 있어요.",
+			"gpt54": "Poe 제공업체: Poe를 AI 제공업체로 추가해 Roo Code에서 Poe 모델에 직접 접근할 수 있게 했어요.",
+			"slashSkills": "xAI 및 MiniMax 개선: xAI 제공업체를 Responses API로 마이그레이션하고, 재사용 가능한 변환 유틸리티를 추가하고, Grok-4.20 기본값을 업데이트하고, MiniMax 모델 목록과 컨텍스트 윈도우 처리를 수정해 더 안정적인 설정을 제공해요.",
 			"geminiPro": "Gemini 3.1 Pro 지원: Gemini 3.1 Pro 모델 지원이 추가되었으며, 향상된 성능을 위해 기본 Gemini 모델로 설정되었습니다.",
 			"cliNdjson": "CLI NDJSON 프로토콜: 새로운 NDJSON stdin 프로토콜, list 하위 명령어, 더 유연한 CLI 워크플로를 위한 모듈화된 run 명령어.",
 			"cliRelease": "CLI v0.1.0: Roo Code CLI가 안정적인 명령어 인터페이스로 첫 공식 릴리스에 도달했습니다."
--- a/webview-ui/src/i18n/locales/ko/settings.json
+++ b/webview-ui/src/i18n/locales/ko/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Claude Sonnet 4.x / Claude Opus 4.6의 컨텍스트 창을 100만 토큰으로 확장",
 		"basetenApiKey": "Baseten API 키",
 		"getBasetenApiKey": "Baseten API 키 가져오기",
+		"poeApiKey": "Poe API 키",
+		"getPoeApiKey": "Poe API 키 가져오기",
+		"poeBaseUrl": "Poe 기본 URL",
 		"fireworksApiKey": "Fireworks API 키",
 		"getFireworksApiKey": "Fireworks API 키 받기",
 		"deepSeekApiKey": "DeepSeek API 키",
--- a/webview-ui/src/i18n/locales/nl/chat.json
+++ b/webview-ui/src/i18n/locales/nl/chat.json
@ -321,8 +321,8 @@
 		},
 		"release": {
 			"heading": "Wat is er nieuw:",
-			"gpt54": "Ondersteuning voor OpenAI GPT-5.4: OpenAI GPT-5.4 en GPT-5.3 Chat Latest zijn toegevoegd, zodat je de nieuwste OpenAI-chatmodellen in Roo Code kunt gebruiken.",
-			"slashSkills": "Skills als slash-commands: Skills kunnen nu als slash-commands worden aangeboden met fallback-uitvoering voor snellere workflows.",
+			"gpt54": "Poe-provider: Poe toegevoegd als AI-provider, zodat je in Roo Code direct toegang hebt tot Poe-modellen.",
+			"slashSkills": "xAI- en MiniMax-verbeteringen: De xAI-provider gemigreerd naar de Responses API, herbruikbare transformatiehulpmiddelen toegevoegd, Grok-4.20-standaardinstellingen bijgewerkt en MiniMax-modellijsten plus de afhandeling van het contextvenster gerepareerd voor een betrouwbaardere configuratie.",
 			"geminiPro": "Gemini 3.1 Pro ondersteuning: Ondersteuning voor het Gemini 3.1 Pro model is toegevoegd en ingesteld als standaard Gemini-model voor betere prestaties.",
 			"cliNdjson": "CLI NDJSON-protocol: Nieuw NDJSON stdin-protocol, list-subopdrachten en gemodulariseerd run-commando voor flexibelere CLI-workflows.",
 			"cliRelease": "CLI v0.1.0: De Roo Code CLI bereikt zijn eerste officiële release met een stabiele opdrachtinterface."
--- a/webview-ui/src/i18n/locales/nl/settings.json
+++ b/webview-ui/src/i18n/locales/nl/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Breidt het contextvenster uit tot 1 miljoen tokens voor Claude Sonnet 4.x / Claude Opus 4.6",
 		"basetenApiKey": "Baseten API-sleutel",
 		"getBasetenApiKey": "Baseten API-sleutel verkrijgen",
+		"poeApiKey": "Poe API-sleutel",
+		"getPoeApiKey": "Poe API-sleutel verkrijgen",
+		"poeBaseUrl": "Poe Basis-URL",
 		"fireworksApiKey": "Fireworks API-sleutel",
 		"getFireworksApiKey": "Fireworks API-sleutel ophalen",
 		"deepSeekApiKey": "DeepSeek API-sleutel",
--- a/webview-ui/src/i18n/locales/pl/chat.json
+++ b/webview-ui/src/i18n/locales/pl/chat.json
@ -348,8 +348,8 @@
 		},
 		"release": {
 			"heading": "Co nowego:",
-			"gpt54": "Obsługa OpenAI GPT-5.4: Dodaliśmy OpenAI GPT-5.4 i GPT-5.3 Chat Latest, żebyś mógł używać najnowszych modeli czatu OpenAI w Roo Code.",
-			"slashSkills": "Umiejętności jako slash commandy: Skills można teraz udostępniać jako slash commandy z wykonaniem awaryjnym, żeby szybciej uruchamiać workflowy.",
+			"gpt54": "Dostawca Poe: Dodano Poe jako dostawcę AI, żebyś mógł bezpośrednio uzyskać dostęp do modeli Poe w Roo Code.",
+			"slashSkills": "Usprawnienia xAI i MiniMax: Przeniesiono dostawcę xAI do Responses API, dodano narzędzia transformacji wielokrotnego użytku, zaktualizowano domyślne ustawienia Grok-4.20 oraz naprawiono listy modeli MiniMax i obsługę okna kontekstu, żeby konfiguracja była bardziej niezawodna.",
 			"geminiPro": "Wsparcie dla Gemini 3.1 Pro: Dodano wsparcie dla modelu Gemini 3.1 Pro i ustawiono go jako domyślny model Gemini dla lepszej wydajności.",
 			"cliNdjson": "Protokół NDJSON w CLI: Nowy protokół NDJSON przez stdin, podkomendy list oraz zmodularyzowana komenda run dla bardziej elastycznych przepływów pracy CLI.",
 			"cliRelease": "CLI v0.1.0: CLI Roo Code osiąga swoją pierwszą oficjalną wersję ze stabilnym interfejsem poleceń."
--- a/webview-ui/src/i18n/locales/pl/settings.json
+++ b/webview-ui/src/i18n/locales/pl/settings.json
@ -380,6 +380,9 @@
 		"vertex1MContextBetaDescription": "Rozszerza okno kontekstowe do 1 miliona tokenów dla Claude Sonnet 4.x / Claude Opus 4.6",
 		"basetenApiKey": "Klucz API Baseten",
 		"getBasetenApiKey": "Uzyskaj klucz API Baseten",
+		"poeApiKey": "Klucz API Poe",
+		"getPoeApiKey": "Uzyskaj klucz API Poe",
+		"poeBaseUrl": "Bazowy URL Poe",
 		"fireworksApiKey": "Klucz API Fireworks",
 		"getFireworksApiKey": "Uzyskaj klucz API Fireworks",
 		"deepSeekApiKey": "Klucz API DeepSeek",
--- a/Show more
+++ b/Show more