--- title: "v1.82.0 - Realtime Guardrails, Projects Management, and 10+ Performance Optimizations" slug: "v1-82-0" date: 2026-02-28T00:00:00 authors: - name: Krrish Dholakia title: CEO, LiteLLM url: https://www.linkedin.com/in/krish-d/ image_url: https://pbs.twimg.com/profile_images/1298587542745358340/DZv3Oj-h_400x400.jpg - name: Ishaan Jaff title: CTO, LiteLLM url: https://www.linkedin.com/in/reffajnaahsi/ image_url: https://pbs.twimg.com/profile_images/1613813310264340481/lz54oEiB_400x400.jpg hide_table_of_contents: false --- ## Deploy this version import Tabs from '@theme/Tabs'; import TabItem from '@theme/TabItem'; ``` showLineNumbers title="docker run litellm" docker run \ -e STORE_MODEL_IN_DB=True \ -p 4000:4000 \ ghcr.io/berriai/litellm:main-1.82.0-stable ``` ``` showLineNumbers title="pip install litellm" pip install litellm==1.82.0 ``` ## Key Highlights - **Realtime API guardrails** — [Full guardrails support for `/v1/realtime` WebSocket sessions with pre/post-call enforcement, voice transcription hooks, session termination policies, and Vertex AI Gemini Live support](../../docs/proxy/guardrails) - [PR #22152](https://github.com/BerriAI/litellm/pull/22152), [PR #22153](https://github.com/BerriAI/litellm/pull/22153), [PR #22161](https://github.com/BerriAI/litellm/pull/22161), [PR #22165](https://github.com/BerriAI/litellm/pull/22165) - **Projects Management** — [New Projects UI with full CRUD, project-scoped virtual keys, and admin opt-in toggle — organize teams and keys by project](../../docs/proxy/ui_store_model_db_setting) - [PR #22315](https://github.com/BerriAI/litellm/pull/22315), [PR #22360](https://github.com/BerriAI/litellm/pull/22360), [PR #22373](https://github.com/BerriAI/litellm/pull/22373), [PR #22412](https://github.com/BerriAI/litellm/pull/22412) - **Guardrail ecosystem expansion** — [Noma v2, Lakera v2 post-call, Singapore regulatory policies (PDPA + MAS), employment discrimination blockers, code execution blocker, guardrail policy versioning, and production monitoring](../../docs/proxy/guardrails) - [PR #21400](https://github.com/BerriAI/litellm/pull/21400), [PR #21783](https://github.com/BerriAI/litellm/pull/21783), [PR #21948](https://github.com/BerriAI/litellm/pull/21948) - **OpenAI Codex 5.3 — day 0** — [Full support for `gpt-5.3-codex` on OpenAI and Azure, plus `gpt-audio-1.5` and `gpt-realtime-1.5` model coverage](../../docs/providers/openai) - [PR #22035](https://github.com/BerriAI/litellm/pull/22035) - **10+ performance optimizations** — Streaming hot-path fixes, Redis pipeline batching, database task batching, ModelResponse init skip, and router cache improvements — lower latency and CPU on every request - **`/v1/messages` → `/responses` routing** — `/v1/messages` requests are now routed to the [Responses API](../../docs/response_api) by default for OpenAI/Azure models :::danger v1/messages routing change This version starts routing `/v1/messages` requests to the `/responses` API by default. To opt out and continue using chat/completions, set `LITELLM_USE_CHAT_COMPLETIONS_URL_FOR_ANTHROPIC_MESSAGES=true` or `litellm_settings.use_chat_completions_url_for_anthropic_messages: true` in your config. ::: --- ## New Models / Updated Models #### New Model Support (20 new models) | Provider | Model | Context Window | Input ($/1M tokens) | Output ($/1M tokens) | Features | | -------- | ----- | -------------- | ------------------- | -------------------- | -------- | | OpenAI | `gpt-5.3-codex` | 272K | $1.75 | $14.00 | Reasoning, coding | | Azure OpenAI | `azure/gpt-5.3-codex` | 272K | $1.75 | $14.00 | Azure deployment | | OpenAI | `gpt-audio-1.5` | 128K | $2.50 | $10.00 | Audio model | | Azure OpenAI | `azure/gpt-audio-1.5-2026-02-23` | 128K | $2.50 | $10.00 | Audio model | | OpenAI | `gpt-realtime-1.5` | 32K | $4.00 | $16.00 | Realtime model | | Azure OpenAI | `azure/gpt-realtime-1.5-2026-02-23` | 32K | $4.00 | $16.00 | Realtime model | | Groq | `groq/openai/gpt-oss-safeguard-20b` | 131K | $0.075 | $0.30 | Guardrail inference | | Google Vertex AI | `vertex_ai/gemini-3.1-flash-image-preview` | - | - | - | Image generation | | Perplexity | `perplexity/perplexity/sonar` | - | - | - | Sonar search | | Perplexity | `perplexity/openai/gpt-5.1` | - | - | - | Hosted routing | | Perplexity | `perplexity/openai/gpt-5-mini` | - | - | - | Hosted routing | | Perplexity | `perplexity/google/gemini-2.5-flash` | - | - | - | Hosted routing | | Perplexity | `perplexity/google/gemini-2.5-pro` | - | - | - | Hosted routing | | Perplexity | `perplexity/google/gemini-3-flash-preview` | - | - | - | Hosted routing | | Perplexity | `perplexity/google/gemini-3-pro-preview` | - | - | - | Hosted routing | | Perplexity | `perplexity/anthropic/claude-haiku-4-5` | - | - | - | Hosted routing | | Perplexity | `perplexity/anthropic/claude-sonnet-4-5` | - | - | - | Hosted routing | | Perplexity | `perplexity/anthropic/claude-opus-4-5` | - | - | - | Hosted routing | | Perplexity | `perplexity/anthropic/claude-opus-4-6` | - | - | - | Hosted routing | | Perplexity | `perplexity/xai/grok-4-1-fast-non-reasoning` | - | - | - | Hosted routing | #### Features - **[OpenAI](../../docs/providers/openai)** - Day 0 support for `gpt-5.3-codex` on OpenAI and Azure - [PR #22035](https://github.com/BerriAI/litellm/pull/22035) - Add `gpt-audio-1.5` model cost map - [PR #22303](https://github.com/BerriAI/litellm/pull/22303) - Add `gpt-realtime-1.5` model cost map - [PR #22304](https://github.com/BerriAI/litellm/pull/22304) - Add `audio` as supported OpenAI param - [PR #22092](https://github.com/BerriAI/litellm/pull/22092) - Add `prompt_cache_key` and `prompt_cache_retention` support - [PR #20397](https://github.com/BerriAI/litellm/pull/20397) - **[Azure OpenAI](../../docs/providers/azure)** - New Azure OpenAI models 2026-02-25 - [PR #22114](https://github.com/BerriAI/litellm/pull/22114) - **[Anthropic](../../docs/providers/anthropic)** - Add v1 Anthropic Responses API transformation - [PR #22087](https://github.com/BerriAI/litellm/pull/22087) - Sanitize `tool_use` IDs in `convert_to_anthropic_tool_invoke` - [PR #21964](https://github.com/BerriAI/litellm/pull/21964) - Fix model wildcard access issue - [PR #21917](https://github.com/BerriAI/litellm/pull/21917) - **[AWS Bedrock](../../docs/providers/bedrock)** - Encode model ARNs for OpenAI-compatible Bedrock imported models - [PR #21701](https://github.com/BerriAI/litellm/pull/21701) - Support optional regional STS endpoint in role assumption - [PR #21640](https://github.com/BerriAI/litellm/pull/21640) - Native structured outputs API support - [PR #21222](https://github.com/BerriAI/litellm/pull/21222) - **[Google Vertex AI](../../docs/providers/vertex)** - Add `gemini-3.1-flash-image-preview` to model cost map - [PR #22223](https://github.com/BerriAI/litellm/pull/22223) - Enable `context-1m-2025-08-07` beta header for Vertex AI provider - [PR #21867](https://github.com/BerriAI/litellm/pull/21867) - **[OpenRouter](../../docs/providers/openrouter)** - Add OpenRouter native models to model cost map - [PR #20520](https://github.com/BerriAI/litellm/pull/20520) - Add OpenRouter Opus 4.6 to model map - [PR #20525](https://github.com/BerriAI/litellm/pull/20525) - **[Mistral](../../docs/providers/mistral)** - Adjust `mistral-small-2503` input/output cost per token - [PR #22097](https://github.com/BerriAI/litellm/pull/22097) - **[Groq](../../docs/providers/groq)** - Add `groq/openai/gpt-oss-safeguard-20b` model pricing - [PR #21951](https://github.com/BerriAI/litellm/pull/21951) - **[AI/ML](../../docs/providers/aiml)** - Update AIML model pricing - [PR #22139](https://github.com/BerriAI/litellm/pull/22139) - **[Ollama](../../docs/providers/ollama)** - Thread `api_base` to `get_model_info` + graceful fallback - [PR #21970](https://github.com/BerriAI/litellm/pull/21970) - **[PublicAI](../../docs/providers/openai)** - Fix function calling for PublicAI Apertus models - [PR #21582](https://github.com/BerriAI/litellm/pull/21582) - **[xAI](../../docs/providers/xai)** - Add deprecation dates for `grok-2-vision-1212` and `grok-3-mini` models - [PR #20102](https://github.com/BerriAI/litellm/pull/20102) - **General** - Forward auth headers of provider - [PR #22070](https://github.com/BerriAI/litellm/pull/22070) - Normalize camelCase `thinking` param keys to snake_case - [PR #21762](https://github.com/BerriAI/litellm/pull/21762) - Allow `dimensions` param passthrough for non-text-embedding-3 OpenAI models - [PR #22144](https://github.com/BerriAI/litellm/pull/22144) ### Bug Fixes - **[AWS Bedrock](../../docs/providers/bedrock)** - Fix converse handling for `parallel_tool_calls` - [PR #22267](https://github.com/BerriAI/litellm/pull/22267) - Restore `parallel_tool_calls` mapping in `map_openai_params` - [PR #22333](https://github.com/BerriAI/litellm/pull/22333) - Correct `modelInput` format for Converse API batch models - [PR #21656](https://github.com/BerriAI/litellm/pull/21656) - Prevent double UUID in `create_file` S3 key - [PR #21650](https://github.com/BerriAI/litellm/pull/21650) - Filter internal `json_tool_call` when mixed with real tools - [PR #21107](https://github.com/BerriAI/litellm/pull/21107) - Pass timeout param to Bedrock rerank HTTP client - [PR #22021](https://github.com/BerriAI/litellm/pull/22021) - **[Anthropic](../../docs/providers/anthropic)** - Fix model cost map for anthropic fast and `inference_geo` - [PR #21904](https://github.com/BerriAI/litellm/pull/21904) - **[Image Generation](../../docs/image_generation)** - Propagate `extra_headers` to upstream image generation - [PR #22026](https://github.com/BerriAI/litellm/pull/22026) - Add `ChatCompletionImageObject` in `OpenAIChatCompletionAssistantMessage` - [PR #22155](https://github.com/BerriAI/litellm/pull/22155) - **General** - Preserve forwarding of server-side called tools - [PR #22260](https://github.com/BerriAI/litellm/pull/22260) - Fix free model handling from UI paths - [PR #22258](https://github.com/BerriAI/litellm/pull/22258) - Fix `None` TypeError in mapping - [PR #22080](https://github.com/BerriAI/litellm/pull/22080) --- ## LLM API Endpoints #### Features - **[Realtime API](../../docs/response_api)** - Guardrails support for `/v1/realtime` WebSocket endpoint - [PR #22152](https://github.com/BerriAI/litellm/pull/22152) - Vertex AI Gemini Live via unified `/realtime` endpoint - [PR #22153](https://github.com/BerriAI/litellm/pull/22153) - Guardrails with `pre_call`/`post_call` mode on realtime WebSocket - [PR #22161](https://github.com/BerriAI/litellm/pull/22161) - `end_session_after_n_fails` + Endpoint Settings wizard step - [PR #22165](https://github.com/BerriAI/litellm/pull/22165) - Guardrail hook for voice transcription - [PR #21976](https://github.com/BerriAI/litellm/pull/21976) - Fix guardrails not firing for Gemini/Vertex AI and `provider_config` realtime sessions - [PR #22168](https://github.com/BerriAI/litellm/pull/22168) - Add logging, spend tracking support + tool tracing - [PR #22105](https://github.com/BerriAI/litellm/pull/22105) - **[Video Generation](../../docs/video_generation)** - Add `variant` parameter to video content download - [PR #21955](https://github.com/BerriAI/litellm/pull/21955) - Pass `api_key` from `litellm_params` to video remix handlers - [PR #21965](https://github.com/BerriAI/litellm/pull/21965) - Apply custom video pricing from deployment `model_info` - [PR #21923](https://github.com/BerriAI/litellm/pull/21923) - Fix passing of image and parameters in videos API - [PR #22170](https://github.com/BerriAI/litellm/pull/22170) - **[OCR](../../docs/providers/openai#ocr--document-understanding)** - Enable local file support for OCR - [PR #22133](https://github.com/BerriAI/litellm/pull/22133) - **[Websearch / Tool Calling](../../docs/completion/input)** - Preserve thinking blocks in agentic loop follow-up messages - [PR #21604](https://github.com/BerriAI/litellm/pull/21604) - **General** - Add configurable upper bound for chunk processing time - [PR #22209](https://github.com/BerriAI/litellm/pull/22209) - Emit `x-litellm-overhead-duration-ms` header for streaming requests - [PR #22027](https://github.com/BerriAI/litellm/pull/22027) #### Bugs - **General** - Fix mypy attr-defined errors on realtime websocket calls - [PR #22202](https://github.com/BerriAI/litellm/pull/22202) --- ## Management Endpoints / UI #### Features - **Projects** - Add Projects page with list and create flows - [PR #22315](https://github.com/BerriAI/litellm/pull/22315) - Add Project Details page with edit modal - [PR #22360](https://github.com/BerriAI/litellm/pull/22360) - Add project keys table and project dropdown on key create/edit - [PR #22373](https://github.com/BerriAI/litellm/pull/22373) - Add delete project action to Projects table - [PR #22412](https://github.com/BerriAI/litellm/pull/22412) - Add Projects Opt-In Toggle in Admin Settings - [PR #22416](https://github.com/BerriAI/litellm/pull/22416) - Include `created_at` and `updated_at` in `/project/list` response - [PR #22323](https://github.com/BerriAI/litellm/pull/22323) - Add tags in project - [PR #22216](https://github.com/BerriAI/litellm/pull/22216) - **Virtual Keys + Access Groups** - Add bidirectional team/key sync for Access Group CRUD flows - [PR #22253](https://github.com/BerriAI/litellm/pull/22253) - Add pagination and search to `/key/aliases` to prevent OOMs - [PR #22137](https://github.com/BerriAI/litellm/pull/22137) - Add paginated key alias selector in UI - [PR #22157](https://github.com/BerriAI/litellm/pull/22157) - Add `project_id` and `access_group_id` filters for key list endpoint - [PR #22356](https://github.com/BerriAI/litellm/pull/22356) - Add KeyInfoHeader component - [PR #22047](https://github.com/BerriAI/litellm/pull/22047) - Restrict Edit Settings to key owners - [PR #21985](https://github.com/BerriAI/litellm/pull/21985) - Fix virtual key grace period from env/UI - [PR #20321](https://github.com/BerriAI/litellm/pull/20321) - **Agents** - Assign virtual keys to agents - [PR #22045](https://github.com/BerriAI/litellm/pull/22045) - Assign tools to agents - [PR #22064](https://github.com/BerriAI/litellm/pull/22064) - Ensure internal users cannot create agents (RBAC enforcement) - [PR #22329](https://github.com/BerriAI/litellm/pull/22329) - **Proxy Auth / SSO** - OIDC discovery URLs, roles array handling, and dot-notation error hints - [PR #22336](https://github.com/BerriAI/litellm/pull/22336) - Add PROXY_ADMIN role to system user for key rotation - [PR #21896](https://github.com/BerriAI/litellm/pull/21896) - **Usage / Spend Logs** - Add user filtering to usage page - [PR #22059](https://github.com/BerriAI/litellm/pull/22059) - Allow using AI to understand usage patterns - [PR #22042](https://github.com/BerriAI/litellm/pull/22042) - Use backend `request_duration_ms` and make Duration sortable in Logs - [PR #22122](https://github.com/BerriAI/litellm/pull/22122) - Add `request_duration_ms` to SpendLogs - [PR #22066](https://github.com/BerriAI/litellm/pull/22066) - Enrich failure spend logs with key/team metadata - [PR #22049](https://github.com/BerriAI/litellm/pull/22049) - Show real tool names in logs for Anthropic-format tools - [PR #22048](https://github.com/BerriAI/litellm/pull/22048) - **Models + Endpoints** - Show proxy URL in ModelHub - [PR #21660](https://github.com/BerriAI/litellm/pull/21660) - Add `/public/endpoints` for provider endpoint support - [PR #22248](https://github.com/BerriAI/litellm/pull/22248) - **UI Improvements** - Add custom favicon support - [PR #21653](https://github.com/BerriAI/litellm/pull/21653) - Add Blog Dropdown in Navbar - [PR #21859](https://github.com/BerriAI/litellm/pull/21859) - Add UI banner warning for detailed debug mode - [PR #21527](https://github.com/BerriAI/litellm/pull/21527) - Make auth value optional for MCP Server create flow - [PR #22119](https://github.com/BerriAI/litellm/pull/22119) - Tool policies: auto-discover tools + policy enforcement guardrail - [PR #22041](https://github.com/BerriAI/litellm/pull/22041) - **Health Checks** - Add health check max tokens configuration - [PR #22299](https://github.com/BerriAI/litellm/pull/22299) - Limit concurrent health checks with `health_check_concurrency` - [PR #20584](https://github.com/BerriAI/litellm/pull/20584) - Fix health check `model_id` filtering - [PR #21071](https://github.com/BerriAI/litellm/pull/21071) #### Bugs - Populate `user_id` and `user_info` for admin users in `/user/info` - [PR #22239](https://github.com/BerriAI/litellm/pull/22239) - Fix virtual keys pagination stale totals when filtering - [PR #22222](https://github.com/BerriAI/litellm/pull/22222) - Fix Spend Update Queue aggregation never triggers with default presets - [PR #21963](https://github.com/BerriAI/litellm/pull/21963) - Fix timezone config lookup and replace hardcoded timezone map with `ZoneInfo` - [PR #21754](https://github.com/BerriAI/litellm/pull/21754) - Fix custom auth budget issue - [PR #22164](https://github.com/BerriAI/litellm/pull/22164) - Fix missing OAuth session state - [PR #21992](https://github.com/BerriAI/litellm/pull/21992) - Fix Transport Type for OpenAPI Spec on UI - [PR #22005](https://github.com/BerriAI/litellm/pull/22005) - Fix Claude Code plugin schema - [PR #22271](https://github.com/BerriAI/litellm/pull/22271) - Add missing migration for `LiteLLM_ClaudeCodePluginTable` - [PR #22335](https://github.com/BerriAI/litellm/pull/22335) - Only tag selected deployment in access group creation - [PR #21655](https://github.com/BerriAI/litellm/pull/21655) - State management fixes for CheckBatchCost - [PR #21921](https://github.com/BerriAI/litellm/pull/21921) - Remove duplicate antd import in ToolPolicies - [PR #22107](https://github.com/BerriAI/litellm/pull/22107) --- ## AI Integrations ### Logging - **[DataDog](../../docs/proxy/logging#datadog)** - Add ability to trace metrics in DataDog - [PR #22103](https://github.com/BerriAI/litellm/pull/22103) - Correlate LiteLLM call IDs with DataDog APM spans - [PR #22219](https://github.com/BerriAI/litellm/pull/22219) - Fix TTS metric emission issues - [PR #20632](https://github.com/BerriAI/litellm/pull/20632) - **[Prometheus](../../docs/proxy/logging#prometheus)** - Add opt-in `stream` label on `litellm_proxy_total_requests_metric` - [PR #22023](https://github.com/BerriAI/litellm/pull/22023) - Fix team `+Inf` budgets in Prometheus metrics - [PR #22243](https://github.com/BerriAI/litellm/pull/22243) - **[Langfuse](../../docs/proxy/logging#langfuse)** - Fix Langfuse OTEL trace issues - [PR #21309](https://github.com/BerriAI/litellm/pull/21309) - **[Arize Phoenix](../../docs/observability/arize_phoenix)** - Fix nested traces coexistence with OTEL callback - [PR #22169](https://github.com/BerriAI/litellm/pull/22169) - **[Slack](../../docs/proxy/alerting)** - Add optional digest mode for Slack alert types - [PR #21683](https://github.com/BerriAI/litellm/pull/21683) - **General** - Fix Gemini trace ID missing in logging - [PR #22077](https://github.com/BerriAI/litellm/pull/22077) - Populate `cache_read_input_tokens` from `prompt_tokens_details` for OpenAI/Azure - [PR #22090](https://github.com/BerriAI/litellm/pull/22090) ### Guardrails - **[Noma](../../docs/proxy/guardrails)** - Noma guardrails v2 based on custom guardrails framework - [PR #21400](https://github.com/BerriAI/litellm/pull/21400) - **[LakeraAI](../../docs/proxy/guardrails)** - Add Lakera v2 post-call hook with fixed PII masking - [PR #21783](https://github.com/BerriAI/litellm/pull/21783) - **[Presidio](../../docs/proxy/guardrails)** - Fix Presidio streaming and false positives - [PR #21949](https://github.com/BerriAI/litellm/pull/21949) - Fix Presidio streaming v3 reliability improvements - [PR #22283](https://github.com/BerriAI/litellm/pull/22283) - Prevent Presidio crash on non-JSON responses - [PR #22084](https://github.com/BerriAI/litellm/pull/22084) - **Built-in Guardrails** - Block code execution guardrail to prevent agents from executing code - [PR #22154](https://github.com/BerriAI/litellm/pull/22154) - Employment discrimination topic blockers for 5 protected classes - [PR #21962](https://github.com/BerriAI/litellm/pull/21962) - Claims agent guardrails (5 categories + policy template) - [PR #22113](https://github.com/BerriAI/litellm/pull/22113) - New code execution evaluation dataset - [PR #22065](https://github.com/BerriAI/litellm/pull/22065) - Tool policies: auto-discover tools + policy enforcement - [PR #22041](https://github.com/BerriAI/litellm/pull/22041) - **Policy Templates** - Singapore guardrail policies (PDPA + MAS AI Risk Management) - [PR #21948](https://github.com/BerriAI/litellm/pull/21948) - Prefix SG guardrail policy IDs with country code - [PR #21974](https://github.com/BerriAI/litellm/pull/21974) - Guardrail policy versioning - [PR #21862](https://github.com/BerriAI/litellm/pull/21862) - **Guardrail Monitoring** - Guardrail Monitor — measure guardrail reliability in production - [PR #21944](https://github.com/BerriAI/litellm/pull/21944) - **Security** - Fix unauthenticated RCE and sandbox escape in custom code guardrail - [PR #22095](https://github.com/BerriAI/litellm/pull/22095) ### Prompt Management No major prompt management changes in this release. ### Secret Managers No major secret manager changes in this release. --- ## Spend Tracking, Budgets and Rate Limiting - **Priority PayGo cost tracking** for Gemini/Vertex AI - [PR #21909](https://github.com/BerriAI/litellm/pull/21909) - **Add `request_duration_ms` to SpendLogs** for latency tracking per request - [PR #22066](https://github.com/BerriAI/litellm/pull/22066) - **Add `in_flight_requests` metric** to `/health/backlog` + Prometheus - [PR #22319](https://github.com/BerriAI/litellm/pull/22319) - **Enrich failure spend logs** with key/team metadata - [PR #22049](https://github.com/BerriAI/litellm/pull/22049) - **Add spend tracking lifecycle logging** for debugging spend flows - [PR #22029](https://github.com/BerriAI/litellm/pull/22029) - **Fix budget timezone config lookup** and replace hardcoded timezone map with `ZoneInfo` - [PR #21754](https://github.com/BerriAI/litellm/pull/21754) - **Fix Spend Update Queue aggregation** never triggering with default presets - [PR #21963](https://github.com/BerriAI/litellm/pull/21963) - **Avoid mutating caller-owned dicts** in `SpendUpdateQueue` aggregation - [PR #21742](https://github.com/BerriAI/litellm/pull/21742) - **Optimize old spendlog deletion** cron job - [PR #21930](https://github.com/BerriAI/litellm/pull/21930) - **Health check max tokens** configuration - [PR #22299](https://github.com/BerriAI/litellm/pull/22299) --- ## MCP Gateway - **Pass MCP auth headers** from request context to tool fetch for `/v1/responses` and `/chat/completions` - [PR #22291](https://github.com/BerriAI/litellm/pull/22291) - **Default `available_on_public_internet` to true** for MCP server behavior consistency - [PR #22331](https://github.com/BerriAI/litellm/pull/22331) - **Clear error messages** for IP filtering / no available tools - [PR #22142](https://github.com/BerriAI/litellm/pull/22142) - **Strip stale `mcp-session-id` header** to prevent 400 errors across proxy workers - [PR #21417](https://github.com/BerriAI/litellm/pull/21417) - **Skip health check for MCP** with passthrough token auth - [PR #21982](https://github.com/BerriAI/litellm/pull/21982) - **Fix missing OAuth session state** - [PR #21992](https://github.com/BerriAI/litellm/pull/21992) - **Fix Transport Type** for OpenAPI Spec on UI - [PR #22005](https://github.com/BerriAI/litellm/pull/22005) - **Add e2e test** for stateless StreamableHTTP behavior - [PR #22033](https://github.com/BerriAI/litellm/pull/22033) --- ## Performance / Loadbalancing / Reliability improvements **Streaming & hot-path** - Streaming latency improvements — 4 targeted hot-path fixes - [PR #22346](https://github.com/BerriAI/litellm/pull/22346) - Skip throwaway `Usage()` construction in `ModelResponse.__init__` - [PR #21611](https://github.com/BerriAI/litellm/pull/21611) - Optimize `is_model_o_series_model` with `startswith` - [PR #21690](https://github.com/BerriAI/litellm/pull/21690) - Use cached `_safe_get_request_headers` instead of per-request construction - [PR #21430](https://github.com/BerriAI/litellm/pull/21430) - Emit `x-litellm-overhead-duration-ms` header for streaming requests - [PR #22027](https://github.com/BerriAI/litellm/pull/22027) **Database & Redis** - Batch 11 `create_task()` calls into 1 in `update_database()` - [PR #22028](https://github.com/BerriAI/litellm/pull/22028) - Redis pipeline spend updates for batched writes - [PR #22044](https://github.com/BerriAI/litellm/pull/22044) - Recover from prisma-query-engine zombie process - [PR #21899](https://github.com/BerriAI/litellm/pull/21899) - Optimize old spendlog deletion cron job - [PR #21930](https://github.com/BerriAI/litellm/pull/21930) **Router & caching** - Add cache invalidation for `_cached_get_model_group_info` - [PR #20376](https://github.com/BerriAI/litellm/pull/20376) - Remove cache eviction close that kills in-use httpx clients - [PR #22247](https://github.com/BerriAI/litellm/pull/22247) - Store background task references in `LLMClientCache._remove_key` to prevent unawaited coroutine warnings - [PR #22143](https://github.com/BerriAI/litellm/pull/22143) - Fix `ensure_arrival_time` set before calculating queue time - [PR #21918](https://github.com/BerriAI/litellm/pull/21918) **Connection management** - Only set `enable_cleanup_closed` on aiohttp when required - [PR #21897](https://github.com/BerriAI/litellm/pull/21897) - Prometheus child_exit cleanup for gunicorn workers - [PR #22324](https://github.com/BerriAI/litellm/pull/22324) - Prometheus multiprocess cleanup - [PR #22221](https://github.com/BerriAI/litellm/pull/22221) - Limit concurrent health checks with `health_check_concurrency` - [PR #20584](https://github.com/BerriAI/litellm/pull/20584) - Isolate `get_config` failures from model sync loop - [PR #22224](https://github.com/BerriAI/litellm/pull/22224) **Other** - Semantic cache: support configurable vector dimensions - [PR #21649](https://github.com/BerriAI/litellm/pull/21649) - Honor `MAX_STRING_LENGTH_PROMPT_IN_DB` from config env vars - [PR #22106](https://github.com/BerriAI/litellm/pull/22106) - Enhance `MidStreamFallbackError` to preserve original status code and attributes - [PR #22225](https://github.com/BerriAI/litellm/pull/22225) - Network mock utility for testing - [PR #21942](https://github.com/BerriAI/litellm/pull/21942) - Add missing return type annotations to iterator protocol methods in streaming_handler - [PR #21750](https://github.com/BerriAI/litellm/pull/21750) --- ## Security - Fix critical/high CVEs in OS-level libs and NPM transitive dependencies - [PR #22008](https://github.com/BerriAI/litellm/pull/22008) - Fix unauthenticated RCE and sandbox escape in custom code guardrail - [PR #22095](https://github.com/BerriAI/litellm/pull/22095) - Remove hardcoded base64 string flagged by secret scanner - [PR #22125](https://github.com/BerriAI/litellm/pull/22125) --- ## Documentation Updates - Add OpenAI Agents SDK tutorial with LiteLLM Proxy - [PR #21221](https://github.com/BerriAI/litellm/pull/21221) - Add OpenClaw integration tutorial - [PR #21605](https://github.com/BerriAI/litellm/pull/21605) - Add Google GenAI SDK tutorial (JS & Python) - [PR #21885](https://github.com/BerriAI/litellm/pull/21885) - Add Gollem Go agent framework cookbook example - [PR #21747](https://github.com/BerriAI/litellm/pull/21747) - Update AssemblyAI docs with Universal-3 Pro, Speech Understanding, and LLM Gateway - [PR #21130](https://github.com/BerriAI/litellm/pull/21130) - Add `store_model_in_db` release docs - [PR #21863](https://github.com/BerriAI/litellm/pull/21863) - Add Credential Usage Tracking docs - [PR #22112](https://github.com/BerriAI/litellm/pull/22112) - Add proxy request tags docs - [PR #22129](https://github.com/BerriAI/litellm/pull/22129) - Add trailing slash to `/mcp` endpoint URLs - [PR #20509](https://github.com/BerriAI/litellm/pull/20509) - Add pre-PR checklist to UI contributing guide - [PR #21886](https://github.com/BerriAI/litellm/pull/21886) - Replace Azure OpenAI key with mock key in docs - [PR #21997](https://github.com/BerriAI/litellm/pull/21997) - Add performance & reliability section to v1.81.14 release notes - [PR #21950](https://github.com/BerriAI/litellm/pull/21950) - Update v1.81.12-stable release notes to point to stable.1 - [PR #22036](https://github.com/BerriAI/litellm/pull/22036) - Add security vulnerability scan report to v1.81.14 release notes - [PR #22385](https://github.com/BerriAI/litellm/pull/22385) --- ## New Contributors * @janfrederickk made their first contribution in [PR #21660](https://github.com/BerriAI/litellm/pull/21660) * @hztBUAA made their first contribution in [PR #21656](https://github.com/BerriAI/litellm/pull/21656) * @LeeJuOh made their first contribution in [PR #21754](https://github.com/BerriAI/litellm/pull/21754) * @WhoisMonesh made their first contribution in [PR #21750](https://github.com/BerriAI/litellm/pull/21750) * @trevorprater made their first contribution in [PR #21747](https://github.com/BerriAI/litellm/pull/21747) * @edwiniac made their first contribution in [PR #21870](https://github.com/BerriAI/litellm/pull/21870) * @stakeswky made their first contribution in [PR #21867](https://github.com/BerriAI/litellm/pull/21867) * @ta-stripe made their first contribution in [PR #21701](https://github.com/BerriAI/litellm/pull/21701) * @ron-zhong made their first contribution in [PR #21948](https://github.com/BerriAI/litellm/pull/21948) * @Arindam200 made their first contribution in [PR #21221](https://github.com/BerriAI/litellm/pull/21221) * @Canvinus made their first contribution in [PR #21964](https://github.com/BerriAI/litellm/pull/21964) * @nicolopignatelli made their first contribution in [PR #21951](https://github.com/BerriAI/litellm/pull/21951) * @MarshHawk made their first contribution in [PR #20584](https://github.com/BerriAI/litellm/pull/20584) * @gavksingh made their first contribution in [PR #22106](https://github.com/BerriAI/litellm/pull/22106) * @roni-frantchi made their first contribution in [PR #22090](https://github.com/BerriAI/litellm/pull/22090) * @noahnistler made their first contribution in [PR #22133](https://github.com/BerriAI/litellm/pull/22133) * @dylan-duan-aai made their first contribution in [PR #21130](https://github.com/BerriAI/litellm/pull/21130) * @rasmi made their first contribution in [PR #22322](https://github.com/BerriAI/litellm/pull/22322) --- ## Diff Summary ## 02/28/2026 * New Models / Updated Models: 26 * LLM API Endpoints: 14 * Management Endpoints / UI: 38 * AI Integrations: 25 * Spend Tracking, Budgets and Rate Limiting: 10 * MCP Gateway: 8 * Performance / Loadbalancing / Reliability improvements: 22 * Security: 3 * Documentation Updates: 14 --- ## Full Changelog [v1.81.14.rc.1...v1.82.0](https://github.com/BerriAI/litellm/compare/v1.81.14.rc.1...v1.82.0)