--- title: "v1.92.0 - Claude Sonnet 5, Production MCP OAuth & New Providers" url: "/release_notes/v1.92.0/v1-92-0" canonical_url: "https://docs.litellm.ai/release_notes/v1.92.0/v1-92-0" type: "release-notes" last_updated: "2026-10-09" related: - "/release_notes/v1.92.1/v1-92-1" - "/release_notes/v1.91.4/v1-91-4" --- # v1.92.0 - Claude Sonnet 5, Production MCP OAuth & New Providers > Index of all LiteLLM docs: https://docs.litellm.ai/llms.txt ## Deploy this version **Docker** ```bash docker run \ -e STORE_MODEL_IN_DB=True \ -p 4000:4000 \ docker.litellm.ai/berriai/litellm:1.92.0 ``` **Pip** ```bash pip install litellm==1.92.0 ``` :::danger[Breaking Changes] **Setting `permissions` and `allowed_routes` now requires proxy-admin privileges.** These fields are admin-gated across the key, user, and team endpoints (`/key/update`, `/key/regenerate`, `/user/new`, `/user/update`, and bulk key updates), so non-admin callers that previously set them will now be rejected. See [PR #31810](https://github.com/BerriAI/litellm/pull/31810), [PR #31987](https://github.com/BerriAI/litellm/pull/31987). ::: ## Key Highlights - **Claude Sonnet 5** - first-class support across Anthropic, Amazon Bedrock (including the regional inference profiles), Vertex AI, and Azure AI, with a 1M-token context window, reasoning, computer use, PDF input, and introductory pricing through 2026-08-31. - **Production-ready MCP OAuth (On-Behalf-Of)** - the token_exchange arm moves onto the v2 resolver with RFC 9728 -> RFC 8414 endpoint discovery (no IdP guessing), persisted Dynamic Client Registration, per-server outbound concurrency limits, and a `mcp_tool_search` virtual tool for large tool catalogs. - **Two new providers** - Tencent (DeepSeek V4 flash/pro via TokenHub, chat and `/v1/messages`) and Google Distributed Cloud (GDC) Gemini for on-prem/sovereign deployments. - **Access-control hardening** - admin-gating of `permissions` and `allowed_routes` across the key, user, and team endpoints, AES-256-GCM at-rest credential encryption with a versioned re-encryption migration, and redaction of secrets from startup and router error logs. - **Faster hot paths** - spend-counter increments and pre-call budget reads are now gathered concurrently, the cost-callback deepcopy moved off the request event loop, OTel runtime imports are memoized, and Redis-cluster reconnect plus read-replica boot resilience keep the proxy serving during infra blips. ## New Providers and Endpoints ### New Providers (2 new providers) | Provider | Supported LiteLLM Endpoints | Description | | --- | --- | --- | | Tencent (`tencent`) | Chat Completions, `/v1/messages` | Tencent TokenHub provider serving DeepSeek V4 (flash and pro) with reasoning, prompt caching, and Anthropic Messages support - [PR #31903](https://github.com/BerriAI/litellm/pull/31903) | | Google Distributed Cloud - GDC (`gdc`) | Chat Completions | Google Distributed Cloud Gemini provider for on-prem and sovereign-cloud deployments - [PR #31895](https://github.com/BerriAI/litellm/pull/31895) | ## New Models / Updated Models #### New Model Support (13 new pricing entries across 4 models) | Provider | Model | Context Window | Input ($/1M tokens) | Output ($/1M tokens) | Features | | --- | --- | --- | --- | --- | --- | | Anthropic / Bedrock / Vertex / Azure AI | `claude-sonnet-5` | 1M | $2.00 | $10.00 | Reasoning, vision, function calling, prompt caching, computer use, PDF input, adaptive thinking | | Bedrock Mantle | `bedrock_mantle/xai.grok-4.3` | 131K | $1.25 | $2.50 | Reasoning, vision, function calling | | Tencent | `tencent/deepseek-v4-flash` | 1M | $0.14 | $0.28 | Reasoning, function calling, prompt caching | | Tencent | `tencent/deepseek-v4-pro` | 1M | $0.435 | $0.87 | Reasoning, function calling, prompt caching | Claude Sonnet 5 ships with pricing entries for Anthropic (`claude-sonnet-5`), Amazon Bedrock (`anthropic.claude-sonnet-5` plus the `us` / `eu` / `au` / `jp` / `global` regional inference profiles), Vertex AI (`vertex_ai/claude-sonnet-5`), and Azure AI (`azure_ai/claude-sonnet-5`) - [PR #31740](https://github.com/BerriAI/litellm/pull/31740), with introductory pricing applied through 2026-08-31 - [PR #31917](https://github.com/BerriAI/litellm/pull/31917). #### Features - **[Anthropic](../../docs/providers/anthropic)** - Add Claude Sonnet 5 with reasoning, computer use, and PDF input - [PR #31740](https://github.com/BerriAI/litellm/pull/31740) - Keep `context_management` working when `drop_params` is enabled - [PR #32020](https://github.com/BerriAI/litellm/pull/32020) - Preserve `x-anthropic-billing-header` system blocks for first-party Anthropic - [PR #29584](https://github.com/BerriAI/litellm/pull/29584) - **[Amazon Bedrock](../../docs/providers/bedrock)** - Forward `strict` and `additionalProperties` to the Converse `toolSpec` - [PR #29814](https://github.com/BerriAI/litellm/pull/29814) - Add `xai.grok-4.3` to the model cost map for Bedrock Mantle SigV4 auth - [PR #31916](https://github.com/BerriAI/litellm/pull/31916) - SigV4/IAM auth on the Bedrock Mantle Responses API route - [PR #29788](https://github.com/BerriAI/litellm/pull/29788) - Honor `ttl` for `tool_config` cache injection points - [PR #31929](https://github.com/BerriAI/litellm/pull/31929) - Map `guardrailConfig` to InvokeModel guardrail headers - [PR #31985](https://github.com/BerriAI/litellm/pull/31985) - **[Google Vertex AI](../../docs/providers/vertex)** - Pass the full `imageConfig` dict for Gemini image generation - [PR #31811](https://github.com/BerriAI/litellm/pull/31811) - Propagate Vertex AI metadata in streaming success callbacks - [PR #29899](https://github.com/BerriAI/litellm/pull/29899) - Single media upload for batch files to fix 499s on large uploads - [PR #31653](https://github.com/BerriAI/litellm/pull/31653) - **[Google Gemini](../../docs/providers/gemini)** - Support Gemini TTS `languageCode` parameters - [PR #29623](https://github.com/BerriAI/litellm/pull/29623) - New Google Distributed Cloud (GDC) Gemini provider - [PR #31895](https://github.com/BerriAI/litellm/pull/31895) - **[Tencent](../../docs/providers)** - Add Tencent TokenHub as a provider serving DeepSeek V4 - [PR #31903](https://github.com/BerriAI/litellm/pull/31903) - **[Fireworks AI](../../docs/providers/fireworks_ai)** - Enable tool calling for `glm-5p1` in the model cost map - [PR #29697](https://github.com/BerriAI/litellm/pull/29697) - **[Databricks](../../docs/providers/databricks)** - Split parallel tool calls so each tool message follows its `tool_calls` - [PR #31633](https://github.com/BerriAI/litellm/pull/31633) - **General** - Add Parasail as a JSON-configured OpenAI-compatible provider - [PR #29842](https://github.com/BerriAI/litellm/pull/29842) ### Bug Fixes - **[Amazon Bedrock](../../docs/providers/bedrock)** - Drop `toolSpec.strict` for Opus 4.7/4.8 to unblock tool calls - [PR #31923](https://github.com/BerriAI/litellm/pull/31923) - Drop `strict`/`additionalProperties` from `toolSpec` for Claude Sonnet 4 - [PR #31943](https://github.com/BerriAI/litellm/pull/31943) - Omit empty `additionalModelRequestFields` and `system` from the Converse payload - [PR #29565](https://github.com/BerriAI/litellm/pull/29565) - Expand `os.environ/` references for AWS auth params, then all fields, in DB-sourced models - [PR #32256](https://github.com/BerriAI/litellm/pull/32256), [PR #32405](https://github.com/BerriAI/litellm/pull/32405) - **[Anthropic](../../docs/providers/anthropic)** - Bill streaming 1h prompt-cache writes at the 1h rate - [PR #32073](https://github.com/BerriAI/litellm/pull/32073) - Drop unsignable thinking blocks and allow a null signature in logging - [PR #31654](https://github.com/BerriAI/litellm/pull/31654) - Require a caller `api_key` and validate `api_base` in the advisor tool - [PR #32093](https://github.com/BerriAI/litellm/pull/32093) ## LLM API Endpoints #### Features - **[Responses API](../../docs/response_api)** - Passthrough `/v1/messages` to native provider endpoints via `supported_endpoints` - [PR #31685](https://github.com/BerriAI/litellm/pull/31685) - Route GitHub Copilot `/v1/messages` to the native Anthropic endpoint - [PR #31802](https://github.com/BerriAI/litellm/pull/31802) - Add cache-control injection support for the `/v1/messages` endpoint - [PR #31778](https://github.com/BerriAI/litellm/pull/31778) - **[/v1/messages](../../docs/anthropic_unified)** - Drop top-level `additional_drop_params` on `/v1/messages` - [PR #31645](https://github.com/BerriAI/litellm/pull/31645) - **[Image Generation](../../docs/image_generation)** - Azure AI MAI-Image-2.5 image generation support - [PR #29688](https://github.com/BerriAI/litellm/pull/29688) - **[A2A](../../docs/a2a)** - Support `a2a-sdk` 1.x proxy routing for 0.3 and 1.0 agents - [PR #30950](https://github.com/BerriAI/litellm/pull/30950) - **[Sandbox](../../docs/sandbox)** - Reuse the e2b container across requests when `metadata.session_id` is set - [PR #31688](https://github.com/BerriAI/litellm/pull/31688) - **General** - Extend the response-headers hook to streaming, TTS, image generation, and pass-through - [PR #24232](https://github.com/BerriAI/litellm/pull/24232) - `websearch_interception` agentic-loop fixes for chat completions and Anthropic messages - [PR #31669](https://github.com/BerriAI/litellm/pull/31669) - Make the TinyFish search provider permissive and attribute errors - [PR #31997](https://github.com/BerriAI/litellm/pull/31997) #### Bugs - **[Responses API](../../docs/response_api)** - Preserve forced-function `tool_choice` name in the Responses-to-Chat transform - [PR #29812](https://github.com/BerriAI/litellm/pull/29812) - Map a system-only chat request to a system input item in the Responses bridge - [PR #29817](https://github.com/BerriAI/litellm/pull/29817) - Merge `metadata.tags` into `litellm_metadata` on the `/v1/responses` route - [PR #31793](https://github.com/BerriAI/litellm/pull/31793) - Route per-model on GitHub Copilot `/v1/responses` based on model info - [PR #29747](https://github.com/BerriAI/litellm/pull/29747) - Drop unmappable Bedrock Responses tools instead of failing the request - [PR #31663](https://github.com/BerriAI/litellm/pull/31663) - **[Realtime API](../../docs/realtime)** - Stop a second Gemini Live setup, retry a hung handshake, and close a guardrail bypass - [PR #31519](https://github.com/BerriAI/litellm/pull/31519) - Trigger Nova Sonic generation on `response.create` so realtime sessions stop hanging - [PR #31924](https://github.com/BerriAI/litellm/pull/31924) - Route realtime HTTP endpoints through the router for credential resolution - [PR #32077](https://github.com/BerriAI/litellm/pull/32077) - **[OCR](../../docs/ocr)** - Preserve content, tables, and `keyValuePairs` in Azure AI doc-intelligence `/v1/ocr` - [PR #32018](https://github.com/BerriAI/litellm/pull/32018) - **[A2A](../../docs/a2a)** - Populate response usage in the A2A chat transformation - [PR #31980](https://github.com/BerriAI/litellm/pull/31980) - **General** - Return upstream error bodies unchanged in pass-through - [PR #32133](https://github.com/BerriAI/litellm/pull/32133) - Normalize Anthropic pass-through server tool usage - [PR #29827](https://github.com/BerriAI/litellm/pull/29827) - Redact provider credentials embedded in fallback error messages - [PR #32083](https://github.com/BerriAI/litellm/pull/32083) - Support Responses input in the Redis semantic cache - [PR #29581](https://github.com/BerriAI/litellm/pull/29581) ## Management Endpoints / UI #### Features - **Virtual Keys & Access Control** - Admin-gate `permissions` on `/key/update`, `/key/regenerate`, `/user/new`, `/user/update`, and bulk key updates - [PR #31810](https://github.com/BerriAI/litellm/pull/31810), [PR #31998](https://github.com/BerriAI/litellm/pull/31998), [PR #32002](https://github.com/BerriAI/litellm/pull/32002) - Admin-gate `allowed_routes` presence on `/key/update` and `/key/regenerate` - [PR #31987](https://github.com/BerriAI/litellm/pull/31987) - Gate non-admin `/key/generate` `budget_limits` and permissions - [PR #31469](https://github.com/BerriAI/litellm/pull/31469) - Reject team-scoped `object_permission` on personal keys for non-admins - [PR #31471](https://github.com/BerriAI/litellm/pull/31471) - Support `object_permission` in `default_key_generate_params` - [PR #31776](https://github.com/BerriAI/litellm/pull/31776) - Reject non-finite `budget_limits` windows and hard-reject CLI session-token personal-key budgets on `/key/generate` - [PR #31630](https://github.com/BerriAI/litellm/pull/31630), [PR #31631](https://github.com/BerriAI/litellm/pull/31631) - Tighten role gating on the `/get/config/callbacks` response and extend banned-params + admin-clear lists - [PR #31745](https://github.com/BerriAI/litellm/pull/31745), [PR #31742](https://github.com/BerriAI/litellm/pull/31742) - Audit default-user-settings and remaining system-wide-settings updates - [PR #31753](https://github.com/BerriAI/litellm/pull/31753), [PR #31754](https://github.com/BerriAI/litellm/pull/31754) - JWT auth opt-in fallback to the DB team on an unresolved claim - [PR #28913](https://github.com/BerriAI/litellm/pull/28913) - Reject non-existent team/key/model scope entries on policy attachment create - [PR #32131](https://github.com/BerriAI/litellm/pull/32131) - **UI** - shadcn migration foundation: Tailwind v4, shadcn init, and antd cascade fix - [PR #31995](https://github.com/BerriAI/litellm/pull/31995) - Migrate the chat UI from antd to shadcn/ui and add key-management and usage panels - [PR #32074](https://github.com/BerriAI/litellm/pull/32074) - Rotate model credentials in a dedicated modal so a normal save can't overwrite secrets - [PR #28089](https://github.com/BerriAI/litellm/pull/28089) - Disclaim that the Update API Key modal only rotates `api_key` - [PR #31805](https://github.com/BerriAI/litellm/pull/31805) - Add budget duration to the edit-team-member form - [PR #29717](https://github.com/BerriAI/litellm/pull/29717) - Render provider icons on the public model hub - [PR #29958](https://github.com/BerriAI/litellm/pull/29958) #### Bugs - **Virtual Keys & Models** - Show team projects to internal users on key creation - [PR #28855](https://github.com/BerriAI/litellm/pull/28855) - Label the default key type as "Full Access" on the key edit page - [PR #29870](https://github.com/BerriAI/litellm/pull/29870) - Keep virtual-keys filters across delete and refresh - [PR #31533](https://github.com/BerriAI/litellm/pull/31533) - Allow deleting a BYOK model after its team is deleted, and delete a team's BYOK models on team deletion - [PR #29875](https://github.com/BerriAI/litellm/pull/29875), [PR #29977](https://github.com/BerriAI/litellm/pull/29977) - Count only legacy `function_call.arguments` in the token counter - [PR #31741](https://github.com/BerriAI/litellm/pull/31741) - Fix the typo `generic_role_mappoings` -> `generic_role_mappings` - [PR #29753](https://github.com/BerriAI/litellm/pull/29753) - **UI** - Stop the Request Logs page from overflowing horizontally and size its columns - [PR #31426](https://github.com/BerriAI/litellm/pull/31426) - Fix the Router Settings Loadbalancing tab save - [PR #31735](https://github.com/BerriAI/litellm/pull/31735) - Allow any git host on the skills add form - [PR #31652](https://github.com/BerriAI/litellm/pull/31652) - Include cache token columns in the usage export - [PR #32015](https://github.com/BerriAI/litellm/pull/32015) - Unify migrated-route URLs and migrate the API Reference page - [PR #29953](https://github.com/BerriAI/litellm/pull/29953) - Make the workflow runs page fill full width - [PR #29868](https://github.com/BerriAI/litellm/pull/29868) ## AI Integrations ### Logging - **[Prometheus](../../docs/proxy/prometheus)** - Add an `api_provider` label to token, latency, request, and cache metrics - [PR #32126](https://github.com/BerriAI/litellm/pull/32126) - Add `litellm_overhead_with_guardrails_latency_metric` - [PR #31593](https://github.com/BerriAI/litellm/pull/31593) - Expose `project_alias` in custom metadata labels and expose MCP tool metadata - [PR #31784](https://github.com/BerriAI/litellm/pull/31784), [PR #31899](https://github.com/BerriAI/litellm/pull/31899) - Bound per-request budget metric emission with a timeout - [PR #31632](https://github.com/BerriAI/litellm/pull/31632) - **[OpenTelemetry](../../docs/observability/opentelemetry_integration)** - Stamp `gen_ai.response.time_to_first_chunk` on streaming LLM spans - [PR #32236](https://github.com/BerriAI/litellm/pull/32236) - Restore `error.*` span attributes on v2 error spans - [PR #32524](https://github.com/BerriAI/litellm/pull/32524) - **[S3](../../docs/proxy/logging)** - Send `Content-MD5` on PUT and support optional server-side encryption in the s3 v2 logger - [PR #31928](https://github.com/BerriAI/litellm/pull/31928) - **[Microsoft Sentinel](../../docs/proxy/logging)** - Resolve the audit stream from `AZURE_SENTINEL_AUDIT_STREAM_NAME` - [PR #32010](https://github.com/BerriAI/litellm/pull/32010) - **[FOCUS Export](../../docs/proxy/cost_tracking)** - Include organization metadata in Vantage FOCUS Tags export and add a GCS destination for FOCUS export - [PR #28184](https://github.com/BerriAI/litellm/pull/28184), [PR #29751](https://github.com/BerriAI/litellm/pull/29751) - **General** - Route realtime success logging through the bounded worker - [PR #31733](https://github.com/BerriAI/litellm/pull/31733) - Restore the admin key/team `callback_vars.turn_off_message_logging` override - [PR #31905](https://github.com/BerriAI/litellm/pull/31905) - Resolve `model_map_value` for proxy custom pricing in standard logging - [PR #31940](https://github.com/BerriAI/litellm/pull/31940) - Log hashed cache keys - [PR #29890](https://github.com/BerriAI/litellm/pull/29890) - Add a Galileo health check for the UI callback test - [PR #29908](https://github.com/BerriAI/litellm/pull/29908) ### Guardrails - **[Model Armor](../../docs/proxy/guardrails/model_armor)** - Scan file and document attachments with Model Armor - [PR #31655](https://github.com/BerriAI/litellm/pull/31655) - **[CrowdStrike AIDR](/docs/proxy/guardrails/quick_start)** - Capture user and model metadata, reading identity from both metadata bags - [PR #29517](https://github.com/BerriAI/litellm/pull/29517), [PR #29991](https://github.com/BerriAI/litellm/pull/29991) - **[Headroom](/docs/proxy/guardrails/quick_start)** - Add CCR (compress-cache-retrieve) via an agentic loop - [PR #31681](https://github.com/BerriAI/litellm/pull/31681) - Add an `unreachable_fallback` fail-open option to the headroom guardrail - [PR #32026](https://github.com/BerriAI/litellm/pull/32026) - **General** - Buffer streamed responses until moderated, with a clean SSE on block - [PR #31389](https://github.com/BerriAI/litellm/pull/31389) - Expose streaming knobs on `generic_guardrail_api` - [PR #31730](https://github.com/BerriAI/litellm/pull/31730) - Keep the create-guardrail modal open on outside click and default the guardrails page to the Guardrails tab - [PR #29871](https://github.com/BerriAI/litellm/pull/29871), [PR #29872](https://github.com/BerriAI/litellm/pull/29872) ### Secret Managers - **General** - AES-256-GCM at-rest credential encryption with a versioned format and a re-encryption migration - [PR #31215](https://github.com/BerriAI/litellm/pull/31215) ## Spend Tracking, Budgets and Rate Limiting - **Budgets & Fallbacks** - Key-level `budget_fallbacks` to reroute requests when a per-model budget is exceeded - [PR #31783](https://github.com/BerriAI/litellm/pull/31783), with UI configuration on the key create/edit forms - [PR #32072](https://github.com/BerriAI/litellm/pull/32072) - Add a `disable_budget_reservation` general setting - [PR #29493](https://github.com/BerriAI/litellm/pull/29493) - Reserve team-budget raises for proxy admins and don't block `/team/update` on an unchanged budget - [PR #30030](https://github.com/BerriAI/litellm/pull/30030), [PR #29525](https://github.com/BerriAI/litellm/pull/29525) - Prevent duplicate budget alert emails on concurrent threshold crossings and apply `EMAIL_SIGNATURE` to them - [PR #32011](https://github.com/BerriAI/litellm/pull/32011), [PR #31712](https://github.com/BerriAI/litellm/pull/31712) - **Cost Tracking** - Standardize rate-limit errors with `category`, `rate_limit_type`, `model`, and `llm_provider` fields - [PR #27687](https://github.com/BerriAI/litellm/pull/27687) - Store the cost breakdown for `/v1/realtime` sessions - [PR #30069](https://github.com/BerriAI/litellm/pull/30069) - Track cost for unmanaged Vertex AI batch jobs - [PR #31442](https://github.com/BerriAI/litellm/pull/31442) - Report real token usage on blocked responses - [PR #31217](https://github.com/BerriAI/litellm/pull/31217) - Emit the `x-litellm-response-cost` header on `/messages` and `/generateContent` - [PR #31675](https://github.com/BerriAI/litellm/pull/31675) - Recognize `*.cognitiveservices.azure.com` as OpenAI-compatible in pass-through cost tracking - [PR #29730](https://github.com/BerriAI/litellm/pull/29730) - Record agent `cost_per_query` and input tokens on the A2A native send path - [PR #31979](https://github.com/BerriAI/litellm/pull/31979) - Count only active users toward the license seat limit - [PR #31227](https://github.com/BerriAI/litellm/pull/31227) - Log per-token-type reasoning and cache cost breakdown - [PR #31623](https://github.com/BerriAI/litellm/pull/31623) ## MCP Gateway - **OAuth 2.0 (On-Behalf-Of) v2** - Migrate the token_exchange (OBO) arm to the v2 resolver and make it production-ready with discovery threading, audit hardening, and an RFC 9728 challenge - [PR #31526](https://github.com/BerriAI/litellm/pull/31526), [PR #31622](https://github.com/BerriAI/litellm/pull/31622) - Discover the OBO token endpoint via RFC 9728 -> RFC 8414 instead of guessing the IdP - [PR #31762](https://github.com/BerriAI/litellm/pull/31762) - Persist the DCR `client_id` so interactive OAuth token refresh works, including on-create Authorize & Fetch - [PR #31912](https://github.com/BerriAI/litellm/pull/31912), [PR #31920](https://github.com/BerriAI/litellm/pull/31920) - Resolve per-user OAuth identity authoritatively at the token endpoint - [PR #31657](https://github.com/BerriAI/litellm/pull/31657) - Support `client_secret_basic` for upstream OAuth token endpoints and add a token-endpoint auth-method selector in the UI - [PR #31635](https://github.com/BerriAI/litellm/pull/31635), [PR #31739](https://github.com/BerriAI/litellm/pull/31739) - Gate OAuth authorize/token/register/discovery on `auth_type=oauth2` - [PR #31736](https://github.com/BerriAI/litellm/pull/31736) - Mirror the upstream token lifetime instead of forcing a 1h OBO expiry - [PR #29951](https://github.com/BerriAI/litellm/pull/29951) - Reset OAuth state on create-server modal close so a prior server's token no longer leaks into the next add-server session - [PR #30000](https://github.com/BerriAI/litellm/pull/30000) - Let non-creator users OAuth into OBO-mode servers and allow team access-group grants in the authorize/token check - [PR #29867](https://github.com/BerriAI/litellm/pull/29867), [PR #30041](https://github.com/BerriAI/litellm/pull/30041) - **Server Management & Tools** - Add `mcp_tool_search` virtual tools for large tool catalogs - [PR #31777](https://github.com/BerriAI/litellm/pull/31777) - Bound outbound tool-call concurrency per MCP server - [PR #31641](https://github.com/BerriAI/litellm/pull/31641) - Add an `all-proxy-mcpservers` sentinel to grant teams every MCP server - [PR #32012](https://github.com/BerriAI/litellm/pull/32012) - Roll up MCP tool spend to user counters and the usage UI - [PR #31576](https://github.com/BerriAI/litellm/pull/31576) - Hydrate the MCP server registry from the DB on startup when `store_model_in_db` is false - [PR #31775](https://github.com/BerriAI/litellm/pull/31775) - Emit a `tools/list` CLIENT span for MCP discovery under otel_v2 - [PR #31525](https://github.com/BerriAI/litellm/pull/31525) - **Bug Fixes** - Stop one unauthenticated server from emptying the aggregate `tools/list` - [PR #31684](https://github.com/BerriAI/litellm/pull/31684) - Surface `tools/list` auth failures as a 401 challenge on single-server routes - [PR #31921](https://github.com/BerriAI/litellm/pull/31921) - Load MCP tool configuration tools via the OBO/passthrough-aware GET path - [PR #29960](https://github.com/BerriAI/litellm/pull/29960) - Tighten role-based visibility on `/v1/mcp/server/submissions` - [PR #31932](https://github.com/BerriAI/litellm/pull/31932) - Highlight MCP cards red when the logged-in user is missing per-user env vars - [PR #29856](https://github.com/BerriAI/litellm/pull/29856) - BYOM visibility, preview UX, and admin-settings gating - [PR #31809](https://github.com/BerriAI/litellm/pull/31809) - Re-add the chat UI and allow a simple UI for MCP OBO auth - [PR #31893](https://github.com/BerriAI/litellm/pull/31893) - Keep an in-flight OAuth resume from resetting when the create-server modal mounts closed - [PR #32416](https://github.com/BerriAI/litellm/pull/32416) ## Performance / Loadbalancing / Reliability improvements - **Spend & Auth hot paths** - Gather independent per-scope spend-counter increments - [PR #31578](https://github.com/BerriAI/litellm/pull/31578) - Move the cost-callback payload deepcopy off the request event loop - [PR #31579](https://github.com/BerriAI/litellm/pull/31579) - Gather independent pre-call budget-enforcement reads in `common_checks` - [PR #31604](https://github.com/BerriAI/litellm/pull/31604) - Memoize the per-request lazy import of OTel runtime hooks - [PR #31707](https://github.com/BerriAI/litellm/pull/31707) - Load the virtual-keys team filter from the fast v2 endpoint - [PR #31638](https://github.com/BerriAI/litellm/pull/31638) - **Reliability** - Keep serving reads from the read replica when the primary DB is down at startup - [PR #31951](https://github.com/BerriAI/litellm/pull/31951) - Re-establish async Redis-cluster connections after a node restart - [PR #31577](https://github.com/BerriAI/litellm/pull/31577) - Isolate poison spend-log rows so one bad record can't drop the whole batch - [PR #31705](https://github.com/BerriAI/litellm/pull/31705) - Stop leaking `master_key` and `database_url` in startup DEBUG logs - [PR #31944](https://github.com/BerriAI/litellm/pull/31944) - **Routing** - Tag-routing denylist support via a `!` prefix - [PR #31728](https://github.com/BerriAI/litellm/pull/31728) - Declarative fallback generalizations for unknown models - [PR #29718](https://github.com/BerriAI/litellm/pull/29718) - Skip the health check for semantic auto_router deployments - [PR #31668](https://github.com/BerriAI/litellm/pull/31668) - **Build** - Bump the wolfi-base digest for glibc 2.43-r10 - [PR #32277](https://github.com/BerriAI/litellm/pull/32277) ## Documentation Updates - Require a reproduction video for reported issues in the contribution guidelines - [PR #30063](https://github.com/BerriAI/litellm/pull/30063) ### PR roll-up by ownership area PRs by ownership area (total: 226) - LLM API Endpoints: 30 - UI: 28 - MCP: 28 - Models & Providers: 27 - Other (CI / chore / tests / version bumps): 24 - Auth & Management: 23 - Logging: 22 - Spend / Budgets / Rate Limits: 18 - Guardrails: 12 - Performance: 11 - Docs: 2 - Secret Managers: 1 ## New Contributors - @roytev made their first contribution in [PR #29565](https://github.com/BerriAI/litellm/pull/29565) - @balcsida made their first contribution in [PR #29581](https://github.com/BerriAI/litellm/pull/29581) - @PigeonMark made their first contribution in [PR #29584](https://github.com/BerriAI/litellm/pull/29584) - @johngarrido made their first contribution in [PR #29623](https://github.com/BerriAI/litellm/pull/29623) - @arnav-144p made their first contribution in [PR #29753](https://github.com/BerriAI/litellm/pull/29753) - @Kaihuang724 made their first contribution in [PR #29842](https://github.com/BerriAI/litellm/pull/29842) - @fengjikui made their first contribution in [PR #29890](https://github.com/BerriAI/litellm/pull/29890) - @fernando-izar made their first contribution in [PR #31632](https://github.com/BerriAI/litellm/pull/31632) ## Full Changelog [`v1.91.0...v1.92.0`](https://github.com/BerriAI/litellm/compare/v1.91.0...v1.92.0) ## Related pages - [v1.92.1](https://docs.litellm.ai/release_notes/v1.92.1/v1-92-1.md) - [v1.91.4](https://docs.litellm.ai/release_notes/v1.91.4/v1-91-4.md)