--- title: "v1.85.0 - Realtime GA, MCP Gateway Expansion & Hardened Multi-Tenancy" url: "/release_notes/v1.85.0/v1-85-0" canonical_url: "https://docs.litellm.ai/release_notes/v1.85.0/v1-85-0" type: "release-notes" last_updated: "2026-10-07" related: - "/release_notes/v1.85.1/v1-85-1" - "/release_notes/v1.84.10/v1-84-10" --- # v1.85.0 - Realtime GA, MCP Gateway Expansion & Hardened Multi-Tenancy > 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:1.85.0 ``` **Pip** ```bash pip install litellm==1.85.0 ``` ## Key Highlights - **OpenAI Realtime GA** — first-class support for the GA OpenAI Realtime API (plus beta compatibility), including `gpt-realtime-2` pricing and `/openai/v1/realtime` logging. - **Hardened multi-tenancy** — a large sweep of per-tenant scoping fixes across keys, projects, batches, files, MCP servers, and analytics endpoints (project-hijack/key-org isolation, service-account resource isolation, per-entity team/agent activity scoping). - **MCP Gateway expansion** — org-level MCP server/toolset permissions, OBO (on-behalf-of) MCP auth, `delegate_auth_to_upstream` PKCE passthrough, and MCP access-group name namespacing. - **Observability overhaul** — broad Prometheus fixes (label-count correctness, end-user cardinality cap, PromQL escaping), OTEL handler isolation + GenAI message-content capture, and decoupled S3 audit-log config. - **New models** — xAI `grok-4.3` / `grok-4.3-latest`, OpenAI `gpt-realtime-2`, OpenRouter `qwen/qwen3.6-plus`, SambaNova `MiniMax-M2.7`, and Bedrock Z.AI `GLM-5`. --- ## New Models / Updated Models #### New Model Support (5 new models) | Provider | Model | Context Window | Input ($/1M tokens) | Output ($/1M tokens) | Features | | --- | --- | --- | --- | --- | --- | | OpenAI | `gpt-realtime-2` | 32K | $4.00 (audio in $32.00) | $16.00 (audio out $64.00) | Realtime (`/v1/realtime`), audio in/out, function calling, parallel tool calls | | xAI | `xai/grok-4.3` | 1M | $1.25 (>200K: $2.50) | $2.50 (>200K: $5.00) | Reasoning, vision, prompt caching, response schema, web search, tool calling | | xAI | `xai/grok-4.3-latest` | 1M | $1.25 (>200K: $2.50) | $2.50 (>200K: $5.00) | Reasoning, vision, prompt caching, response schema, web search, tool calling | | OpenRouter | `openrouter/qwen/qwen3.6-plus` | 1M | $0.325 | $1.95 | Reasoning, vision, function calling, tool choice | | SambaNova | `sambanova/MiniMax-M2.7` | 204.8K | $0.30 | $1.20 | Reasoning, function calling, tool choice | Pricing/metadata also updated for existing entries: Gemini multimodal-embedding pricing repointed to the Vertex pricing source with image/audio/video per-unit costs, audio-token cost reductions on realtime/Gemini entries, and a `gemini-embedding-2-preview` cost alignment. - xAI grok-4.3 / grok-4.3-latest metadata - [PR #27154](https://github.com/BerriAI/litellm/pull/27154), [PR #27396](https://github.com/BerriAI/litellm/pull/27396) - OpenAI gpt-realtime-2 pricing - [PR #27653](https://github.com/BerriAI/litellm/pull/27653) - OpenRouter Qwen 3.6 Plus metadata - [PR #27486](https://github.com/BerriAI/litellm/pull/27486) - New chat model metadata + Bedrock Z.AI GLM-5 - [PR #27313](https://github.com/BerriAI/litellm/pull/27313), [PR #24338](https://github.com/BerriAI/litellm/pull/24338) - GPT-4o-Transcribe pricing fix - [PR #27875](https://github.com/BerriAI/litellm/pull/27875) #### Features - **[Anthropic](../../docs/providers/anthropic)** - Forward `output_config.effort`, reject garbage `reasoning_effort` with 400, and omit thinking/output_config when `reasoning_effort="none"` - [PR #27074](https://github.com/BerriAI/litellm/pull/27074), [PR #27039](https://github.com/BerriAI/litellm/pull/27039) - Add Bedrock Claude Platform route - [PR #27678](https://github.com/BerriAI/litellm/pull/27678) - Inject dummy tool without `modify_params` - [PR #27620](https://github.com/BerriAI/litellm/pull/27620) - **[Bedrock](../../docs/providers/bedrock)** - Add Z.AI GLM-5 model support - [PR #24338](https://github.com/BerriAI/litellm/pull/24338) - Handle document content blocks in Converse API message conversion - [PR #24644](https://github.com/BerriAI/litellm/pull/24644) - Refactor response stream shape handling - [PR #27257](https://github.com/BerriAI/litellm/pull/27257) - **[Vertex AI](../../docs/providers/vertex)** - Model Garden OpenAPI support for publisher model IDs - [PR #26076](https://github.com/BerriAI/litellm/pull/26076) - Omit `system_instruction`/`tools`/`toolConfig` when `cachedContent` set - [PR #26077](https://github.com/BerriAI/litellm/pull/26077) - **[Gemini](../../docs/providers/gemini)** - Follow provider defaults for Gemini 3 thinking - [PR #25764](https://github.com/BerriAI/litellm/pull/25764) - Handle Gemini Files API URIs without fetching - [PR #24922](https://github.com/BerriAI/litellm/pull/24922) - Normalize `response_schema` on native `generateContent` - [PR #27775](https://github.com/BerriAI/litellm/pull/27775) - **[xAI](../../docs/providers/xai)** - Add `parallel_tool_calls` to supported params - [PR #25106](https://github.com/BerriAI/litellm/pull/25106) - **[Azure](../../docs/providers/azure)** - Authenticate to Azure with a token - [PR #27556](https://github.com/BerriAI/litellm/pull/27556) - Azure Sentinel audit-log support - [PR #27280](https://github.com/BerriAI/litellm/pull/27280) - **General** - `gpt-5.5` reasoning-effort capability flags + `supports_low_reasoning_effort` - [PR #26456](https://github.com/BerriAI/litellm/pull/26456) - Match `litellm.completion` supported params with proxy model info - [PR #27720](https://github.com/BerriAI/litellm/pull/27720) #### Bug Fixes - **[OpenRouter](../../docs/providers/openrouter)** - Strip `openrouter/` prefix from model names - [PR #24282](https://github.com/BerriAI/litellm/pull/24282) - **[Azure](../../docs/providers/azure)** - Forward `api_version` to `aembedding()` for Azure AI Foundry v1 endpoints - [PR #24911](https://github.com/BerriAI/litellm/pull/24911) - Route Azure container file requests by decoded deployment - [PR #26402](https://github.com/BerriAI/litellm/pull/26402) - **[Anthropic](../../docs/providers/anthropic)** / **[Vertex](../../docs/providers/vertex)** - Fix Vertex Anthropic streaming status-error hangs - [PR #27310](https://github.com/BerriAI/litellm/pull/27310) - Fix Anthropic streaming reasoning token usage - [PR #27319](https://github.com/BerriAI/litellm/pull/27319) - **[Fireworks AI](../../docs/providers/fireworks_ai)** - Strip `thinking_blocks` from chat messages before the Fireworks API call - [PR #27881](https://github.com/BerriAI/litellm/pull/27881) - **[hosted vLLM](../../docs/providers/vllm)** - Normalize custom tools for chat completions - [PR #25763](https://github.com/BerriAI/litellm/pull/25763) - **General** - Decode unified `file_id` when `model_file_id_mapping` is unavailable - [PR #27406](https://github.com/BerriAI/litellm/pull/27406) - Pass `output_config` through to backends that accept it - [PR #26439](https://github.com/BerriAI/litellm/pull/26439) - Resolve provider from deployment for multi-provider default config - [PR #27517](https://github.com/BerriAI/litellm/pull/27517) - Return `503` from `/health` when the targeted model is unhealthy or DB is disconnected - [PR #27003](https://github.com/BerriAI/litellm/pull/27003) - Guard URL-valued model destinations and align resource-model auth checks - [PR #26915](https://github.com/BerriAI/litellm/pull/26915), [PR #26963](https://github.com/BerriAI/litellm/pull/26963) ## LLM API Endpoints #### Features - **[Realtime API](../../docs/realtime)** - OpenAI Realtime GA support and beta compatibility - [PR #27110](https://github.com/BerriAI/litellm/pull/27110) - Add `/openai/v1/realtime` to routes for logging - [PR #27323](https://github.com/BerriAI/litellm/pull/27323) - **[Responses API](../../docs/response_api)** - Persist and replay streamed Responses API requests from cache - [PR #24580](https://github.com/BerriAI/litellm/pull/24580) - Route `gpt-5.4+` chat-without-tools to the Responses API - [PR #27618](https://github.com/BerriAI/litellm/pull/27618) - Preserve `cache_control` in Responses → Chat Completion transformation - [PR #27727](https://github.com/BerriAI/litellm/pull/27727) - Normalize chat `tool_choice` for the completions→responses bridge - [PR #27634](https://github.com/BerriAI/litellm/pull/27634) - **[Batches](../../docs/batches)** - Bedrock batch model-invocation job retrieval - [PR #26834](https://github.com/BerriAI/litellm/pull/26834) - Transform Vertex AI batch prediction outputs to OpenAI format - [PR #25627](https://github.com/BerriAI/litellm/pull/25627) - Set `response=null` on batch error entries per OpenAI spec - [PR #27041](https://github.com/BerriAI/litellm/pull/27041) - **[Embeddings](../../docs/embedding/supported_embedding)** - Default OpenAI-path `encoding_format` to `float` - [PR #26976](https://github.com/BerriAI/litellm/pull/26976) - Separate embeddings for multimodal inputs + combined multimodal embeddings via nested input - [PR #24337](https://github.com/BerriAI/litellm/pull/24337), [PR #24341](https://github.com/BerriAI/litellm/pull/24341) - **[Audio Transcription](../../docs/audio_transcription)** - Add NVIDIA Riva STT provider - [PR #27185](https://github.com/BerriAI/litellm/pull/27185) - **[Vector Stores](../../docs/vector_stores)** - Resolve embedding config at request time, never persist credentials - [PR #27082](https://github.com/BerriAI/litellm/pull/27082) - Tighten managed-store access - [PR #26930](https://github.com/BerriAI/litellm/pull/26930) #### Bugs - **General** - Preserve `compact_20260112` context management on Bedrock `/v1/messages` - [PR #27534](https://github.com/BerriAI/litellm/pull/27534) - Fix managed file `model_mappings` when the router resolves a single deployment dict - [PR #26950](https://github.com/BerriAI/litellm/pull/26950) - Omit `model` from Azure deployment image-gen / image-edit bodies - [PR #27103](https://github.com/BerriAI/litellm/pull/27103) - Fix Bedrock passthrough call-ID headers - [PR #27412](https://github.com/BerriAI/litellm/pull/27412) - Pin Responses API affinity to the Azure resource on model-group switch - [PR #27703](https://github.com/BerriAI/litellm/pull/27703) - Align `vertex_ai/gemini-embedding-2-preview` cost with Vertex multimodal pricing - [PR #27848](https://github.com/BerriAI/litellm/pull/27848) - Consolidate batch + dynamic limiter check/increment - [PR #26954](https://github.com/BerriAI/litellm/pull/26954) ## Management Endpoints / UI #### Features - **Virtual Keys** - Bulk key updates for a team - [PR #26468](https://github.com/BerriAI/litellm/pull/26468) - Rename "Default" key type to "Full Access" and reorder dropdown - [PR #27218](https://github.com/BerriAI/litellm/pull/27218) - Add `Expires` to key Overview header; merge User into one field - [PR #27696](https://github.com/BerriAI/litellm/pull/27696) - **Teams & Models** - Search teams by team ID alongside name - [PR #27684](https://github.com/BerriAI/litellm/pull/27684) - Add a "Your Usage" view for admin users on the usage page - [PR #26746](https://github.com/BerriAI/litellm/pull/26746) - Add Vertex AI Search as a vector-store provider in the UI - [PR #27790](https://github.com/BerriAI/litellm/pull/27790) - "Last Minute" quick-select on the Logs time range - [PR #27446](https://github.com/BerriAI/litellm/pull/27446) - Add missing Z.AI (`zai`) provider to the Add-Model dropdown - [PR #26419](https://github.com/BerriAI/litellm/pull/26419) - **SSO / Auth** - JWT scope + wildcard support for routing overrides, issuer verification with unscoped warning - [PR #26325](https://github.com/BerriAI/litellm/pull/26325), [PR #27008](https://github.com/BerriAI/litellm/pull/27008) - Grafana Cloud Pyroscope authentication - [PR #26902](https://github.com/BerriAI/litellm/pull/26902) - Show full IdP claims in `/sso/debug/callback` - [PR #27498](https://github.com/BerriAI/litellm/pull/27498) #### Bugs: access scoping & correctness - **Multi-tenancy isolation** - Scope project, key-org, team, and agent-activity lookups per entity; reject `user_id=None` on non-admin analytics; re-validate `user_id` after `/user/info` re-parses query - [PR #27011](https://github.com/BerriAI/litellm/pull/27011), [PR #27014](https://github.com/BerriAI/litellm/pull/27014), [PR #26929](https://github.com/BerriAI/litellm/pull/26929), [PR #27009](https://github.com/BerriAI/litellm/pull/27009) - Constrain cloud-storage file paths and batch-file model access - [PR #27019](https://github.com/BerriAI/litellm/pull/27019), [PR #27015](https://github.com/BerriAI/litellm/pull/27015) - Isolate managed resources for service-account API keys - [PR #27004](https://github.com/BerriAI/litellm/pull/27004) - Tighten resource-ownership checks and sensitive public-endpoint guards - [PR #26951](https://github.com/BerriAI/litellm/pull/26951), [PR #26912](https://github.com/BerriAI/litellm/pull/26912) - **Authorization hardening** - Block missing write routes for proxy admin viewers; restore admin-viewer read parity on Logs + Settings - [PR #27007](https://github.com/BerriAI/litellm/pull/27007), [PR #26846](https://github.com/BerriAI/litellm/pull/26846) - Encode upstream URL path identifiers; require a trusted proxy for header-identity auth - [PR #26860](https://github.com/BerriAI/litellm/pull/26860), [PR #26825](https://github.com/BerriAI/litellm/pull/26825) - Bind generic SSO state to a session cookie; allow non-admin compliance-path reads - [PR #26944](https://github.com/BerriAI/litellm/pull/26944), [PR #27234](https://github.com/BerriAI/litellm/pull/27234) - **Keys / Teams / SCIM** - Honor `key access_group_ids` when a team restricts models; resolve access-group names in team filtering and same-name deployment routing - [PR #26275](https://github.com/BerriAI/litellm/pull/26275), [PR #25224](https://github.com/BerriAI/litellm/pull/25224), [PR #26161](https://github.com/BerriAI/litellm/pull/26161) - Revoke virtual keys when SCIM deprovisions a user; fix SCIM user-lookup filters - [PR #26861](https://github.com/BerriAI/litellm/pull/26861), [PR #27308](https://github.com/BerriAI/litellm/pull/27308) - Key-rotation bug fix; honor `team_member_permissions` on `/key/list` - [PR #27756](https://github.com/BerriAI/litellm/pull/27756), [PR #27026](https://github.com/BerriAI/litellm/pull/27026) - `/config/update` targeted per-section writes (drop `store_model_in_db` gate) - [PR #26643](https://github.com/BerriAI/litellm/pull/26643) - Scope CLI stored token to `base_url`; redact Gemini API key from URL query params in error traces - [PR #26945](https://github.com/BerriAI/litellm/pull/26945), [PR #24943](https://github.com/BerriAI/litellm/pull/24943) - **UI fixes** - Remove the insecure `?token=` URL handler from the login page; clear admin session cookies before establishing an invited user's session; URL-encode `team_id` in `teamInfoCall` - [PR #26924](https://github.com/BerriAI/litellm/pull/26924), [PR #27227](https://github.com/BerriAI/litellm/pull/27227), [PR #27466](https://github.com/BerriAI/litellm/pull/27466) - Project dropdown empty for internal users (3 bugs); remove blank leading entry from access-group model dropdown; omit `allowed_routes` from key edit save when unchanged - [PR #26664](https://github.com/BerriAI/litellm/pull/26664), [PR #27521](https://github.com/BerriAI/litellm/pull/27521), [PR #27553](https://github.com/BerriAI/litellm/pull/27553) - Member/team access-group fix; team model test-connection authorization - [PR #27317](https://github.com/BerriAI/litellm/pull/27317), [PR #27487](https://github.com/BerriAI/litellm/pull/27487) ## AI Integrations ### Logging - **[Prometheus](../../docs/proxy/prometheus)** - Fix custom-metadata label counts, cap end-user metric cardinality, fix remaining-metric zero values, escape `api_key` for PromQL string literals, emit `litellm_remaining_tokens_metric` for Bedrock & Vertex - [PR #27268](https://github.com/BerriAI/litellm/pull/27268), [PR #27272](https://github.com/BerriAI/litellm/pull/27272), [PR #27348](https://github.com/BerriAI/litellm/pull/27348), [PR #27013](https://github.com/BerriAI/litellm/pull/27013), [PR #27705](https://github.com/BerriAI/litellm/pull/27705) - Fix `/metrics` hang when `require_auth_for_metrics_endpoint` is true and auth succeeds; point `/metrics` 401 at the opt-out flag; fix metric labels for litellm-side rejects - [PR #25980](https://github.com/BerriAI/litellm/pull/25980), [PR #27502](https://github.com/BerriAI/litellm/pull/27502), [PR #26947](https://github.com/BerriAI/litellm/pull/26947) - **[OpenTelemetry](../../docs/observability/opentelemetry_integration)** - Isolate dual OTEL handlers; honor `OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT`; fix proxy-integration tracing bugs - [PR #27018](https://github.com/BerriAI/litellm/pull/27018), [PR #27403](https://github.com/BerriAI/litellm/pull/27403), [PR #27757](https://github.com/BerriAI/litellm/pull/27757) - **[Arize](../../docs/observability/arize_integration)** / **LangSmith** - Arize `_set_usage_outputs` handles raw OpenAI Pydantic `CompletionUsage`; remove unwanted metadata info from LangSmith - [PR #26506](https://github.com/BerriAI/litellm/pull/26506), [PR #26894](https://github.com/BerriAI/litellm/pull/26894) - **General** - Decouple S3 audit-log config via `s3_audit_callback_params` - [PR #27222](https://github.com/BerriAI/litellm/pull/27222) - Set `verbose_logger` level when `LITELLM_LOG=INFO`; require a team-management role on `/team/{id}/callback`; close callback-config and observability-credential side channels; guard dynamic integration hosts - [PR #26401](https://github.com/BerriAI/litellm/pull/26401), [PR #26819](https://github.com/BerriAI/litellm/pull/26819), [PR #27081](https://github.com/BerriAI/litellm/pull/27081), [PR #26921](https://github.com/BerriAI/litellm/pull/26921) ### Guardrails - **General** - Add Qohash Nexus guardrail hook - [PR #24927](https://github.com/BerriAI/litellm/pull/24927) - Run model-level `post_call` guardrails on streaming requests; ensure post-call guardrail fires exactly once - [PR #26922](https://github.com/BerriAI/litellm/pull/26922), [PR #27012](https://github.com/BerriAI/litellm/pull/27012), [PR #26109](https://github.com/BerriAI/litellm/pull/26109) - Preserve Responses event streams in Presidio output masking - [PR #26878](https://github.com/BerriAI/litellm/pull/26878) - Cover multimodal + Responses-API content shapes; tighten tool-permission checks; optional skip of tool message in unified guardrail inputs - [PR #26957](https://github.com/BerriAI/litellm/pull/26957), [PR #26969](https://github.com/BerriAI/litellm/pull/26969), [PR #27441](https://github.com/BerriAI/litellm/pull/27441) - Handle legacy dict shape for `metadata.guardrails` in the Team UI - [PR #27224](https://github.com/BerriAI/litellm/pull/27224) ### Prompt Management - **General** - Block path-traversal in BitBucket / Arize Phoenix / AssemblyAI clients; sandbox jinja2 in the GitLab/Arize/BitBucket prompt managers - [PR #26943](https://github.com/BerriAI/litellm/pull/26943), [PR #27043](https://github.com/BerriAI/litellm/pull/27043) ### Secret Managers - **General** - Audit-log `/cache/settings` and `/config_overrides/hashicorp_vault` mutations - [PR #26953](https://github.com/BerriAI/litellm/pull/26953) ## Spend Tracking, Budgets and Rate Limiting - **Rate Limiting** - Atomic TPM rate limit; include model name + configured TPM/RPM in priority rate-limit 429 errors - [PR #27001](https://github.com/BerriAI/litellm/pull/27001), [PR #27216](https://github.com/BerriAI/litellm/pull/27216) - Load team-member RPM/TPM from membership budget in the combined view - [PR #24925](https://github.com/BerriAI/litellm/pull/24925) - **Budgets** - Skip the personal-budget hook when a reservation covers the counter - [PR #27021](https://github.com/BerriAI/litellm/pull/27021) - Treat `0` `team_member_budget` as no cap; enforce team-member budget without a user row; reset org/tag/proxy budgets correctly - [PR #27133](https://github.com/BerriAI/litellm/pull/27133), [PR #27273](https://github.com/BerriAI/litellm/pull/27273), [PR #27326](https://github.com/BerriAI/litellm/pull/27326), [PR #27488](https://github.com/BerriAI/litellm/pull/27488) - Flush virtual-key `model_max` budget spend to Redis after success logging; tighten budget spend admission - [PR #27334](https://github.com/BerriAI/litellm/pull/27334), [PR #26845](https://github.com/BerriAI/litellm/pull/26845) - **Tag Budgets & Routing** - Enforce tag budgets on `x-litellm-tags` header requests; tag-budget reset drops stale management-cache entries; union `x-litellm-tags` with static team/key tags; fix internal tag-usage scoping; always merge caller-supplied tags into request metadata - [PR #27573](https://github.com/BerriAI/litellm/pull/27573), [PR #27568](https://github.com/BerriAI/litellm/pull/27568), [PR #27247](https://github.com/BerriAI/litellm/pull/27247), [PR #27315](https://github.com/BerriAI/litellm/pull/27315), [PR #27784](https://github.com/BerriAI/litellm/pull/27784) - Tag-routing test preventing header-regex bypass for strict plain-text tags - [PR #26805](https://github.com/BerriAI/litellm/pull/26805) - **Spend Logs / Cost** - Pass `service_tier` through Azure and Azure AI cost calculation - [PR #24926](https://github.com/BerriAI/litellm/pull/24926) - Opt-in suppression of stack traces in spend-tracking error logs; keep spend-log cleanup running after batch failures; redact echoed prompts in `error_information`; prevent `secret_fields` from leaking into spend logs; drop client-supplied pricing fields from request bodies - [PR #26899](https://github.com/BerriAI/litellm/pull/26899), [PR #27303](https://github.com/BerriAI/litellm/pull/27303), [PR #27689](https://github.com/BerriAI/litellm/pull/27689), [PR #27143](https://github.com/BerriAI/litellm/pull/27143), [PR #27071](https://github.com/BerriAI/litellm/pull/27071) ## MCP Gateway - **Features** - Org-level MCP server and toolset permissions - [PR #26960](https://github.com/BerriAI/litellm/pull/26960) - OBO (on-behalf-of) MCP auth - [PR #27421](https://github.com/BerriAI/litellm/pull/27421) - `delegate_auth_to_upstream` flag for PKCE passthrough - [PR #27834](https://github.com/BerriAI/litellm/pull/27834) - Support MCP access-group names in URL-based namespacing - [PR #27726](https://github.com/BerriAI/litellm/pull/27726) - **Bugs** - Sanitize tool names to Anthropic's `[a-zA-Z0-9_-]{1,128}` pattern - [PR #26788](https://github.com/BerriAI/litellm/pull/26788) - Require a trusted-proxy gate before honoring `X-Forwarded-*` on OAuth discovery; preserve oauth2 m2m auth for tools routes; run `pre_call_tool_check` on the OpenAPI/local-registry path - [PR #26841](https://github.com/BerriAI/litellm/pull/26841), [PR #26871](https://github.com/BerriAI/litellm/pull/26871), [PR #27016](https://github.com/BerriAI/litellm/pull/27016) - Redact MCP server URL/headers for non-admin viewers; replace user-API-key auth with authorization-or-cookie for MCP server creation - [PR #27027](https://github.com/BerriAI/litellm/pull/27027), [PR #27190](https://github.com/BerriAI/litellm/pull/27190) - Fix MCP DB reload partial failures; surface upstream 401 for token-forwarding MCP servers - [PR #27314](https://github.com/BerriAI/litellm/pull/27314), [PR #27847](https://github.com/BerriAI/litellm/pull/27847) ## Performance / Loadbalancing / Reliability improvements - **Routing & Reliability** - Trigger fallbacks on mid-stream `httpx.TimeoutException` - [PR #26998](https://github.com/BerriAI/litellm/pull/26998) - Register cooldowns on failure + fail fast on stale `encrypted_content` (Responses) - [PR #27820](https://github.com/BerriAI/litellm/pull/27820) - Register model info under the responses/-stripped variant - [PR #27531](https://github.com/BerriAI/litellm/pull/27531) - Fix Redis Sentinel client handling for authenticated Sentinel setups - [PR #26302](https://github.com/BerriAI/litellm/pull/26302) - **Proxy hot path** - Token-verification query optimization - [PR #26202](https://github.com/BerriAI/litellm/pull/26202) - Run daily activity aggregation off the event loop - [PR #27264](https://github.com/BerriAI/litellm/pull/27264) - Shared IAM cache + static credentials in `BaseAWSLLM` - [PR #27125](https://github.com/BerriAI/litellm/pull/27125) - Isolate semantic cache entries; stable Redis key generation across working directories; remove a duplicate in-memory cache-size constant - [PR #26990](https://github.com/BerriAI/litellm/pull/26990), [PR #27025](https://github.com/BerriAI/litellm/pull/27025), [PR #26385](https://github.com/BerriAI/litellm/pull/26385) - Early proxy request-size enforcement; coerce non-str `x-litellm-*` header values to avoid an httpx `TypeError` - [PR #27311](https://github.com/BerriAI/litellm/pull/27311), [PR #27504](https://github.com/BerriAI/litellm/pull/27504) - Separate DB read and write endpoints - [PR #27493](https://github.com/BerriAI/litellm/pull/27493) - **Health checks** - Shared health-check polling; `health_check_reasoning_effort` for model health checks; skip `disable_background_health_check` models on `GET /health`; scope `/health` response to the caller's models; remove the separate health app - [PR #26434](https://github.com/BerriAI/litellm/pull/26434), [PR #27115](https://github.com/BerriAI/litellm/pull/27115), [PR #27716](https://github.com/BerriAI/litellm/pull/27716), [PR #26935](https://github.com/BerriAI/litellm/pull/26935), [PR #27430](https://github.com/BerriAI/litellm/pull/27430) - **Config / startup robustness** - Hot-reload config YAML when `--reload` is set; break the managed-resources import cycle on Python 3.13; reject bare-str file-input sinks (local-file read hardening) - [PR #27274](https://github.com/BerriAI/litellm/pull/27274), [PR #27160](https://github.com/BerriAI/litellm/pull/27160), [PR #27762](https://github.com/BerriAI/litellm/pull/27762) - **Packaging / Docker / Helm / CI** - Pin Wolfi & uv to multi-arch index digests; remove the hardcoded Prisma binary target for multi-arch builds; clear flagged OS-package advisories on the Docker image; refresh dependency locks - [PR #27123](https://github.com/BerriAI/litellm/pull/27123), [PR #27170](https://github.com/BerriAI/litellm/pull/27170), [PR #27225](https://github.com/BerriAI/litellm/pull/27225), [PR #27126](https://github.com/BerriAI/litellm/pull/27126) - Helm: skip startup `prisma db push` when a migrations Job is enabled; increase default probe timeouts, disable debug logging by default - [PR #27200](https://github.com/BerriAI/litellm/pull/27200), [PR #27237](https://github.com/BerriAI/litellm/pull/27237) - CI: Rerun Failed Tests for all pytest jobs, block PRs that drop coverage, Redis-backed VCR replay caches, reduce cassette bloat, mutation-testing workflow, dev-tag detection in the release workflow, Playwright apt-install skip - [PR #27155](https://github.com/BerriAI/litellm/pull/27155), [PR #27340](https://github.com/BerriAI/litellm/pull/27340), [PR #26838](https://github.com/BerriAI/litellm/pull/26838), [PR #27159](https://github.com/BerriAI/litellm/pull/27159), [PR #27409](https://github.com/BerriAI/litellm/pull/27409), [PR #27576](https://github.com/BerriAI/litellm/pull/27576), [PR #26966](https://github.com/BerriAI/litellm/pull/26966), [PR #27169](https://github.com/BerriAI/litellm/pull/27169) - Remove legacy deployment artifacts and litellm-js packages; remove a redundant backup pricing file; misc test/import cleanup - [PR #27541](https://github.com/BerriAI/litellm/pull/27541), [PR #16590](https://github.com/BerriAI/litellm/pull/16590), [PR #27699](https://github.com/BerriAI/litellm/pull/27699), [PR #27633](https://github.com/BerriAI/litellm/pull/27633) - Tighten router-settings-override and mock-testing trust; drop blank-text fallback for empty Bedrock Converse thinking blocks - [PR #26968](https://github.com/BerriAI/litellm/pull/26968), [PR #27850](https://github.com/BerriAI/litellm/pull/27850) ## Documentation Updates - Update the Greptile README logo to a higher-quality image - [PR #25385](https://github.com/BerriAI/litellm/pull/25385) - Add a `BudgetManager.reset_cost` docstring - [PR #27867](https://github.com/BerriAI/litellm/pull/27867) - Add a `_LoopWrapper` class docstring - [PR #27870](https://github.com/BerriAI/litellm/pull/27870) ## New Contributors - @kimimgo made their first contribution in [#24282](https://github.com/BerriAI/litellm/pull/24282) - @shubham-arora-clear made their first contribution in [#24644](https://github.com/BerriAI/litellm/pull/24644) - @ohnoah made their first contribution in [#24580](https://github.com/BerriAI/litellm/pull/24580) - @ushiromiya-lion made their first contribution in [#25106](https://github.com/BerriAI/litellm/pull/25106) - @gowtham2809 made their first contribution in [#25224](https://github.com/BerriAI/litellm/pull/25224) - @he-yufeng made their first contribution in [#26401](https://github.com/BerriAI/litellm/pull/26401) - @MackDing made their first contribution in [#26419](https://github.com/BerriAI/litellm/pull/26419) - @dgu1-godaddy made their first contribution in [#26834](https://github.com/BerriAI/litellm/pull/26834) - @Vedanshu7 made their first contribution in [#24943](https://github.com/BerriAI/litellm/pull/24943) - @dennishenry made their first contribution in [#27190](https://github.com/BerriAI/litellm/pull/27190) - @SHARP155 made their first contribution in [#27466](https://github.com/BerriAI/litellm/pull/27466) - @mats852 made their first contribution in [#24927](https://github.com/BerriAI/litellm/pull/24927) **Full Changelog**: https://github.com/BerriAI/litellm/compare/v1.84.0...v1.85.0 --- ## 05/16/2026 (`v1.85.0`) > Counts cover PRs **new in `v1.85.0`** relative to `v1.84.0` **stable**. 14 PRs that were backported into `v1.84.0` stable (and documented in the v1.84.0 release notes) are excluded here to avoid double-counting. * New Models / Updated Models: 43 * LLM API Endpoints: 24 * Management Endpoints / UI: 54 * AI Integrations (Logging / Guardrails / Prompt Mgmt / Secret Managers): 32 * Spend Tracking, Budgets and Rate Limiting: 23 * MCP Gateway: 12 * Performance / Loadbalancing / Reliability improvements: 41 * Documentation Updates: 3 Total: 232 PRs --- ## Related pages - [v1.85.1](https://docs.litellm.ai/release_notes/v1.85.1/v1-85-1.md) - [v1.84.10](https://docs.litellm.ai/release_notes/v1.84.10/v1-84-10.md)