v1.111.0-rc.3预发布2026-11-20
Verify this release
``bash
cosign verify \
--certificate-identity https://github.com/BerriAI/project-releaser/.github/workflows/release.yml@refs/heads/main \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
ghcr.io/berriai/litellm:v1.111.0-rc.3
`
Same command for litellm-database, litellm-nonroot and the docker.io/litellm/ images
What's new in 1.111.0-rc.3
15 PRs. 11 fixed. No action needed to upgrade
Changes
- Anthropic: Fixed temperature=1 returning 400 on Claude models that removed sampling params on Anthropic, Vertex AI and Azure AI, broken since v1.110.0-rc.2 (#52341) <sup>Also in 1.110.2</sup>
- Responses API: Fixed streams retrying after output reached the client, which sent duplicate events and billed extra calls (#52461) <sup>Also in 1.110.2</sup>
- Responses API: Fixed streams that fail after reasoning output restarting on a fallback group and sending a second response (#52616) <sup>Also in 1.110.2</sup>
- Caching: Fixed GPT-6.2 system prompts missing the prompt cache on /v1/chat/completions and /v1/messages unless store: false (#52906) <sup>Also in 1.106.12, 1.108.13, 1.109.3, 1.110.2</sup>
- Cost tracking and pricing: Fixed forwarded Anthropic credentials on /v1/messages ignoring the deployment's custom prices, broken since v1.109.0 (#52313) <sup>Also in 1.110.2</sup>
- Keys, teams and auth: Fixed 503s on a burst of first requests for new customer org aliases, which left their spend untracked (#52080)
- Keys, teams and auth: Fixed JWT key mappings without jwtissuer taking 5 minutes to apply when litellmjwtauth.issuers is set, broken since v1.103.0 (#52350) <sup>Also in 1.108.13, 1.109.3, 1.110.2</sup>
- Keys, teams and auth: Fixed /key/info letting a key with no user read other userless keys by hash or alias (#52838, #52925) <sup>Also in 1.106.12, 1.108.13, 1.109.3, 1.110.2</sup>
- Admin UI: Fixed SSO login redirecting to an http://` URL when the proxy runs behind TLS termination, broken since v1.108.0 (#52213)
- Deployment: Fixed each proxy worker opening six extra Postgres connections at boot (#52378)
- Performance: Fixed failed calls holding large request bodies in memory until a full GC, growing workers to several GiB (#52600)
---
Backport #52462 · Backport #52470 · Backport #52691 · Backport #52752 · Backport #52764 · Backport #52914 · Backport #52939 · Backport #53024 · Full changelog
v1.106.122026-11-20
Verify this release
``bash
cosign verify \
--certificate-identity https://github.com/BerriAI/project-releaser/.github/workflows/release.yml@refs/heads/main \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
ghcr.io/berriai/litellm:v1.106.12
`
Same command for litellm-database, litellm-nonroot and the docker.io/litellm/ images
What's new in 1.106.12
6 PRs. 2 fixed, 2 changed. No action needed to upgrade
Changes
- Caching: Fixed GPT-6.2 system prompts missing the prompt cache on /v1/chat/completions and /v1/messages unless store: false (#52906) <sup>Also in 1.108.13, 1.109.3, 1.110.2</sup>
- Keys, teams and auth: Fixed /key/info` letting a key with no user read other userless keys by hash or alias (#52838, #52925) <sup>Also in 1.108.13, 1.109.3, 1.110.2</sup>
- Deployment: Changed dnspython, tornado, filelock, soupsieve and opentelemetry to newer releases (#52936, #53032)
- Deployment: Changed the Docker images to a newer Wolfi base image (#53033)
---
Backport #52924 · Backport #52936 · Full changelog
v1.110.22026-11-20
Verify this release
``bash
cosign verify \
--certificate-identity https://github.com/BerriAI/project-releaser/.github/workflows/release.yml@refs/heads/main \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
ghcr.io/berriai/litellm:v1.110.2
`
Same command for litellm-database, litellm-nonroot and the docker.io/litellm/ images
What's new in 1.110.2
11 PRs. 8 fixed. No action needed to upgrade
Changes
- Anthropic: Fixed temperature=1 returning 400 on Claude models that removed sampling params on Anthropic, Vertex AI and Azure AI, broken since v1.110.0-rc.2 (#52341)
- Responses API: Fixed streams retrying after output reached the client, which sent duplicate events and billed extra calls (#52461)
- Responses API: Fixed streams that fail after reasoning output restarting on a fallback group and sending a second response (#52616)
- Caching: Fixed GPT-6.2 system prompts missing the prompt cache on /v1/chat/completions and /v1/messages unless store: false (#52906) <sup>Also in 1.106.12, 1.108.13, 1.109.3</sup>
- Cost tracking and pricing: Fixed forwarded Anthropic credentials on /v1/messages ignoring the deployment's custom prices, broken since v1.109.0 (#52313)
- Keys, teams and auth: Fixed JWT key mappings without jwtissuer taking 5 minutes to apply when litellmjwtauth.issuers is set, broken since v1.103.0 (#52350) <sup>Also in 1.108.13, 1.109.3</sup>
- Keys, teams and auth: Fixed /key/info` letting a key with no user read other userless keys by hash or alias (#52838, #52925) <sup>Also in 1.106.12, 1.108.13, 1.109.3</sup>
- Performance: Fixed memory growth from every finished request staying in memory until a full GC pass, broken since v1.110.0-rc.1 (#51888)
---
Backport #52463 · Backport #52469 · Backport #52690 · Backport #52751 · Backport #52920 · Backport #52938 · Full changelog
1.108.13 (LTS)2026-11-20
> [!TIP]
> The lts image tag always points to the latest LTS release. Pull ghcr.io/berriai/litellm:lts to stay on the newest LTS patch
Verify this release
``bash
cosign verify \
--certificate-identity https://github.com/BerriAI/project-releaser/.github/workflows/release.yml@refs/heads/main \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
ghcr.io/berriai/litellm:v1.108.13
`
Same command for litellm-database, litellm-nonroot and the docker.io/litellm/ images
What's new in 1.108.13 (LTS)
7 PRs. 5 fixed. No action needed to upgrade
Changes
- Messages API: Fixed Claude Code's rotating billing header breaking prompt caching and inflating input tokens on non-Anthropic models (#52922)
- Caching: Fixed GPT-6.2 system prompts missing the prompt cache on /v1/chat/completions and /v1/messages unless store: false (#52906) <sup>Also in 1.106.12, 1.109.3, 1.110.2</sup>
- Keys, teams and auth: Fixed JWT key mappings without jwtissuer taking 5 minutes to apply when litellmjwtauth.issuers is set, broken since v1.103.0 (#52350) <sup>Also in 1.109.3, 1.110.2</sup>
- Keys, teams and auth: Fixed /key/info letting a key with no user read other userless keys by hash or alias (#52838, #52925) <sup>Also in 1.106.12, 1.109.3, 1.110.2</sup>
- Deployment: Fixed proxy boot on Unix-socket Postgres like Cloud SQL, where the migration lock dialed localhost:5432` (#52797)
---
Backport #52713 · Backport #52819 · Backport #52907 · Backport #52919 · Full changelog
v1.109.32026-11-20
Verify this release
``bash
cosign verify \
--certificate-identity https://github.com/BerriAI/project-releaser/.github/workflows/release.yml@refs/heads/main \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
ghcr.io/berriai/litellm:v1.109.3
`
Same command for litellm-database, litellm-nonroot and the docker.io/litellm/ images
What's new in 1.109.3
6 PRs. 3 fixed, 2 changed. No action needed to upgrade
Changes
- Caching: Fixed GPT-6.2 system prompts missing the prompt cache on /v1/chat/completions and /v1/messages unless store: false (#52906) <sup>Also in 1.106.12, 1.108.13, 1.110.2</sup>
- Keys, teams and auth: Fixed JWT key mappings without jwtissuer taking 5 minutes to apply when litellmjwtauth.issuers is set, broken since v1.103.0 (#52350) <sup>Also in 1.108.13, 1.110.2</sup>
- Keys, teams and auth: Fixed /key/info` letting a key with no user read other userless keys by hash or alias (#52838, #52925) <sup>Also in 1.106.12, 1.108.13, 1.110.2</sup>
- Deployment: Changed filelock to 4.2.4 and soupsieve to 4.1 to match main (#52937)
- Deployment: Changed the Docker images to a newer Wolfi base image (#53008)
---
Backport #52712 · Backport #52929 · Backport #52937 · Full changelog
v1.112.0-dev.1预发布2026-11-20
Verify Docker Image Signature
``bash
cosign verify \
--certificate-identity https://github.com/BerriAI/project-releaser/.github/workflows/release.yml@refs/heads/main \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
ghcr.io/berriai/litellm:v1.112.0-dev.1
``
What's Changed
* test(database): expect no native pool in proxy lifespan cleanup by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/51911
* fix(ui): clarify customer usage periods and preserve date ranges by @moe-berri in https://github.com/BerriAI/litellm/pull/51899
* feat(ui): reuse usage spend design for customer teams and members by @tin-berri in https://github.com/BerriAI/litellm/pull/51724
* feat(customers): move existing members between teams by @moe-berri in https://github.com/BerriAI/litellm/pull/51913
* feat(customers): search customer members by ID or alias in the Members tab by @kenneth-berri in https://github.com/BerriAI/litellm/pull/51896
* test(e2e): drop two batch waits that never finish in time by @yuneng-berri in https://github.com/BerriAI/litellm/pull/51882
* test(integration): record a stack when an owned proxy crashes by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/51922
refactor(s3-log): move the rust-backed s3 logger from s3v3 to s3_v2 by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51893
* perf(host-python): settle ready async values inline instead of spawning on tokio by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51906
* perf(streaming): extract usage once per chunk when assembling stream usage by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51903
* perf(caching): format dual cache debug logs lazily by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51902
perf(auth): read token model fields from the class instead of deprecated fields__ by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51901
* refactor(s3-log): move python-compatible helpers into python-compat by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51895
* fix(proxy): join default teams on the team member budget when the team has one by @kerry-berri in https://github.com/BerriAI/litellm/pull/51728
* fix(proxy): let an exhausted key keep listing models and counting tokens by @kerry-berri in https://github.com/BerriAI/litellm/pull/51675
fix(proxy): run the /falai passthrough price gate on the path sent upstream by @kerry-berri in https://github.com/BerriAI/litellm/pull/51671
fix(falai): send versioned seedream and flux-pro image models to the endpoint they name by @kerry-berri in https://github.com/BerriAI/litellm/pull/51668
* fix(replicate): answer malformed video seconds with a 400 and drop non-decimal size digits instead of a 500 by @kerry-berri in https://github.com/BerriAI/litellm/pull/51670
* fix(perplexity): bill cached prompt tokens at the cache read rate when the response has no cost by @kerry-berri in https://github.com/BerriAI/litellm/pull/51669
fix(vertexai): send extra_body fields as predict parameters for BGE embedding models by @kerry-berri in https://github.com/BerriAI/litellm/pull/51662
fix(bedrock): put extrabody fields inside modelInput for async-invoke embeddings by @kerry-berri in https://github.com/BerriAI/litellm/pull/51676
fix(falai): bill a video request without a resolution at the row's default rate by @kerry-berri in https://github.com/BerriAI/litellm/pull/51665
* fix(router): stop the rust router keeping every finished request in memory until a full gc by @mateo-berri in https://github.com/BerriAI/litellm/pull/51888
* fix(customers): keep uncapped customer spend across scheduled budget resets by @moe-berri in https://github.com/BerriAI/litellm/pull/51918
* fix(budgets): match window starts to stored millisecond precision by @moe-berri in https://github.com/BerriAI/litellm/pull/51921
* fix(ui): default customer usage range to month to date by @moe-berri in https://github.com/BerriAI/litellm/pull/51923
* fix(cost): keep a deployment's per-second video price over the catalog row's flat price by @kerry-berri in https://github.com/BerriAI/litellm/pull/51688
feat(dynamicrate_limiter): customer organizations and keys borrow idle model capacity and split class shares by @kerry-berri in https://github.com/BerriAI/litellm/pull/51705
fix(counttokens): return 400 for malformed Anthropic server-tool blocks by @kerry-berri in https://github.com/BerriAI/litellm/pull/51957
* fix(admin-mcp): restore the budget route operation ids the connector pins by @kerry-berri in https://github.com/BerriAI/litellm/pull/51959
* fix(proxy): hold the logging object weakly in the detached stream failure hook by @kaizen-38 in https://github.com/BerriAI/litellm/pull/51318
* fix(responses): retry a stream that breaks before output in its own model group by @kerry-berri in https://github.com/BerriAI/litellm/pull/51953
* perf(customers): index member team, budget and override lookups by @moe-berri in https://github.com/BerriAI/litellm/pull/51961
* feat(customers): sort organizations, teams and members by budget limit by @moe-berri in https://github.com/BerriAI/litellm/pull/51924
* chore(lint): flag conditionals with more than 3 boolean clauses via ruff PLR0916 by @mateo-berri in https://github.com/BerriAI/litellm/pull/51964
refactor(vertexai): remove unreachable VertexAIFilesHandler.afile_content by @mateo-berri in https://github.com/BerriAI/litellm/pull/51966
* fix(router): free a failed rust router call's request without waiting for a full gc by @mateo-berri in https://github.com/BerriAI/litellm/pull/51963
* perf(auth): reuse validated models on UserApiKeyCache memory hits by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51912
* fix(proxy): reset members given default team member models on the team's budget window by @kerry-berri in https://github.com/BerriAI/litellm/pull/51960
* fix(customers): cascade delete large organizations within the transaction timeout by @moe-berri in https://github.com/BerriAI/litellm/pull/51969
* feat(budgets): replace the python budget hold and settle engine with a rust engine admitted at auth by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51535
fix(dynamicrate_limiter): charge tokens used while the model was idle against a team's class share by @kerry-berri in https://github.com/BerriAI/litellm/pull/51952
* feat(messages): serve bedrock invoke messages on rust only by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51330
* fix(proxy): stop team model writes from pulling in proxy-level settings by @shrey-berri in https://github.com/BerriAI/litellm/pull/51874
* fix(proxy)!: match the sign-in email fallback to the right user by @shrey-berri in https://github.com/BerriAI/litellm/pull/51897
fix(messages): clamp bedrock effort before the dropparams effort gate by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51972
* refactor(proxy): keep the detached stream failure hook state on the logging object by @narmstrong-berri in https://github.com/BerriAI/litellm/pull/51962
* feat(fnmatch): add a crate for Python's fnmatchcase glob dialect by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/51530
* refactor(auth): decide route rules in Rust by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/51531
* refactor(auth): plan the authentication path in Rust by @yujonglee-berri in https://github.com/BerriAI/litellm/pull/51605
fix(decisions): answer 400 naming extraheaders when it is not an object by @paveln83 in https://github.com/BerriAI/litellm/pull/50581
* fix(proxy): stop returning stack traces in health check errors by @ojensen-berri in https://github.com/BerriAI/litellm/pull/52005
chore(cost-map): add falai seedance 2.5 flash rows, copilot deprecation dates, inkling 2 small rows, pixverse clip prices and groq/nebius fixes by @devin-ai-integration[bot] in https://github.com/BerriAI/litellm/pull/51929
* chore(c
1.108.12 (LTS)2026-11-19
> [!TIP]
> The lts image tag always points to the latest LTS release. Pull ghcr.io/berriai/litellm:lts to stay on the newest LTS patch
Verify this release
``bash
cosign verify \
--certificate-identity https://github.com/BerriAI/project-releaser/.github/workflows/release.yml@refs/heads/main \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
ghcr.io/berriai/litellm:v1.108.12
`
Same command for litellm-database, litellm-nonroot and the docker.io/litellm/ images
What's new in 1.108.12 (LTS)
8 PRs. 6 fixed. No action needed to upgrade
Changes
- OpenAI: Fixed GPT-6.2 prompt caching for system prompts sent to /v1/chat/completions or /v1/messages with store: false (#52621, #52739)
- Azure: Fixed azureai/modelrouter on a Foundry /openai/v1 base failing on /v1/responses and /v1/responses/compact (#52474)
- Vertex AI and Gemini: Fixed one json_schema row failing every other row in a Vertex batch file (#51790)
- Messages API: Fixed 400 Tool reference '<name>' not found on /v1/messages when a guardrail sees a tool name over 64 characters (#47434)
- Messages API: Fixed OpenAI models on /v1/messages that keep streaming upstream after the caller closes the stream (#52144)
- Guardrails: Fixed tool_policy letting /v1/messages` requests through when a blocked tool name is over 64 characters (#52717)
---
Backport #52476 · Backport #52666 · Backport #52725 · Full changelog
v1.106.112026-11-19
Verify this release
``bash
cosign verify \
--certificate-identity https://github.com/BerriAI/project-releaser/.github/workflows/release.yml@refs/heads/main \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
ghcr.io/berriai/litellm:v1.106.11
`
Same command for litellm-database, litellm-nonroot and the docker.io/litellm/ images
What's new in 1.106.11
4 PRs. 1 fixed. No action needed to upgrade
Changes
- OpenAI: Fixed GPT-6.2 prompt caching for system prompts sent to /v1/chat/completions or /v1/messages with store: false` (#52599, #52698, #52738)
---
Backport #52665 · Full changelog
v1.111.0-rc.2预发布2026-11-19
Verify this release
``bash
cosign verify \
--certificate-identity https://github.com/BerriAI/project-releaser/.github/workflows/release.yml@refs/heads/main \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
ghcr.io/berriai/litellm:v1.111.0-rc.2
`
Same command for litellm-database, litellm-nonroot and the docker.io/litellm/ images
What's new in 1.111.0-rc.2
1 PRs. 1 fixed. No action needed to upgrade
Changes
- OpenAI: Fixed GPT-6.2 prompt caching for system prompts sent to /v1/chat/completions or /v1/messages with store: false` (#52657)
---
Full changelog
v1.110.12026-11-19
Verify this release
``bash
cosign verify \
--certificate-identity https://github.com/BerriAI/project-releaser/.github/workflows/release.yml@refs/heads/main \
--certificate-oidc-issuer https://token.actions.githubusercontent.com \
ghcr.io/berriai/litellm:v1.110.1
`
Same command for litellm-database, litellm-nonroot and the docker.io/litellm/ images
What's new in 1.110.1
3 PRs. 1 fixed, 2 changed. No action needed to upgrade
Changes
- OpenAI: Fixed GPT-6.2 prompt caching for system prompts sent to /v1/chat/completions or /v1/messages with store: false (#52658)
- Deployment: Changed the UI image to slim nginx, which drops curl but keeps wget and nc` (#52595)
- Deployment: Changed the Docker images to a newer Wolfi base image (#52749)
---
Full changelog