From 9fe06db9216b895c0c1deb6a0b74515b2de083e6 Mon Sep 17 00:00:00 2001
From: "roomote[bot]" <219738659+roomote[bot]@users.noreply.github.com>
Date: Thu, 21 Aug 2025 15:03:35 -0700
Subject: [PATCH] fix: update DeepSeek models context window to 128k (#7269)

* fix: update DeepSeek models context window to 128k

- Updated deepseek-chat and deepseek-reasoner models from 64k to 128k context window
- Updated corresponding test expectations
- Aligns with DeepSeek API documentation at https://api-docs.deepseek.com/quick_start/pricing/

Fixes #7268

* feat: update deepseek-reasoner maxTokens to 64K based on official documentation

* fix: use default maxTokens values instead of maximum for DeepSeek models

- deepseek-chat: 4096 (4K default) instead of 8192 (8K max)
- deepseek-reasoner: 32768 (32K default) instead of 65536 (64K max)
- Updated tests to match new default values
- Updated description to clarify default vs max output tokens

* fix: use maximum output tokens for both DeepSeek models

- deepseek-chat: 8192 (8K max)
- deepseek-reasoner: 65536 (64K max)
- Updated tests to match maximum values
- Updated description to reflect 64K max output

---------

Co-authored-by: Roo Code <roomote@roocode.com>
Co-authored-by: daniel-lxs <ricciodaniel98@gmail.com>
---
 packages/types/src/providers/deepseek.ts     | 10 +++++-----
 src/api/providers/__tests__/deepseek.spec.ts | 18 ++++++++++++++++--
 2 files changed, 21 insertions(+), 7 deletions(-)

diff --git a/packages/types/src/providers/deepseek.ts b/packages/types/src/providers/deepseek.ts
index 5ef757ffdf..8ce5ef87b2 100644
--- a/packages/types/src/providers/deepseek.ts
+++ b/packages/types/src/providers/deepseek.ts
@@ -7,8 +7,8 @@ export const deepSeekDefaultModelId: DeepSeekModelId = "deepseek-chat"
 
 export const deepSeekModels = {
 	"deepseek-chat": {
-		maxTokens: 8192,
-		contextWindow: 64_000,
+		maxTokens: 8192, // 8K max output
+		contextWindow: 128_000,
 		supportsImages: false,
 		supportsPromptCache: true,
 		inputPrice: 0.27, // $0.27 per million tokens (cache miss)
@@ -18,15 +18,15 @@ export const deepSeekModels = {
 		description: `DeepSeek-V3 achieves a significant breakthrough in inference speed over previous models. It tops the leaderboard among open-source models and rivals the most advanced closed-source models globally.`,
 	},
 	"deepseek-reasoner": {
-		maxTokens: 8192,
-		contextWindow: 64_000,
+		maxTokens: 65536, // 64K max output for reasoning mode
+		contextWindow: 128_000,
 		supportsImages: false,
 		supportsPromptCache: true,
 		inputPrice: 0.55, // $0.55 per million tokens (cache miss)
 		outputPrice: 2.19, // $2.19 per million tokens
 		cacheWritesPrice: 0.55, // $0.55 per million tokens (cache miss)
 		cacheReadsPrice: 0.14, // $0.14 per million tokens (cache hit)
-		description: `DeepSeek-R1 achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks. Supports Chain of Thought reasoning with up to 32K tokens.`,
+		description: `DeepSeek-R1 achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks. Supports Chain of Thought reasoning with up to 64K output tokens.`,
 	},
 } as const satisfies Record<string, ModelInfo>
 
diff --git a/src/api/providers/__tests__/deepseek.spec.ts b/src/api/providers/__tests__/deepseek.spec.ts
index 175a5bc44b..50cabfa922 100644
--- a/src/api/providers/__tests__/deepseek.spec.ts
+++ b/src/api/providers/__tests__/deepseek.spec.ts
@@ -154,12 +154,26 @@ describe("DeepSeekHandler", () => {
 			const model = handler.getModel()
 			expect(model.id).toBe(mockOptions.apiModelId)
 			expect(model.info).toBeDefined()
-			expect(model.info.maxTokens).toBe(8192)
-			expect(model.info.contextWindow).toBe(64_000)
+			expect(model.info.maxTokens).toBe(8192) // deepseek-chat has 8K max
+			expect(model.info.contextWindow).toBe(128_000)
 			expect(model.info.supportsImages).toBe(false)
 			expect(model.info.supportsPromptCache).toBe(true) // Should be true now
 		})
 
+		it("should return correct model info for deepseek-reasoner", () => {
+			const handlerWithReasoner = new DeepSeekHandler({
+				...mockOptions,
+				apiModelId: "deepseek-reasoner",
+			})
+			const model = handlerWithReasoner.getModel()
+			expect(model.id).toBe("deepseek-reasoner")
+			expect(model.info).toBeDefined()
+			expect(model.info.maxTokens).toBe(65536) // deepseek-reasoner has 64K max
+			expect(model.info.contextWindow).toBe(128_000)
+			expect(model.info.supportsImages).toBe(false)
+			expect(model.info.supportsPromptCache).toBe(true)
+		})
+
 		it("should return provided model ID with default model info if model does not exist", () => {
 			const handlerWithInvalidModel = new DeepSeekHandler({
 				...mockOptions,