BerriAI/litellm - CodSpeed

litellm

Blog Docs Changelog

Performance History

Latest Results

fix: also correct max_output_tokens for gpt-5-pro (272000→128000) Per reviewer feedback, max_output_tokens was left at 272000 while max_tokens was corrected to 128000, causing an internal inconsistency. Both should be 128000 per OpenAI docs.

xbrxr03:fix/gpt5-context-window-tokens

1 hour ago

fix(logging): dedupe success logging for non-streaming anthropic_messages The /v1/messages (anthropic_messages) route is always dispatched through the async wrapper, which logs via async_success_handler. But unlike acompletion, anthropic_messages sets no async flag in litellm_params, so _is_sync_litellm_request returns True in the sync success_handler and it re-dispatches every CustomLogger. The result for a single non-streaming request was two litellm_request OTEL spans and double cost/success callbacks. Skip the sync CustomLogger pass for anthropic_messages, mirroring how pass_through is already handled, so logging runs once via the canonical async path -- matching acompletion behaviour.

anneheartrecord:fix/anthropic-messages-duplicate-success-logging

3 hours ago

fix(logging): dedupe success logging for non-streaming anthropic_messages The /v1/messages (anthropic_messages) route is always dispatched through the async wrapper, which logs via async_success_handler. But unlike acompletion, anthropic_messages sets no async flag in litellm_params, so _is_sync_litellm_request returns True in the sync success_handler and it re-dispatches every CustomLogger. The result for a single non-streaming request was two litellm_request OTEL spans and double cost/success callbacks. Skip the sync CustomLogger pass for anthropic_messages, mirroring how pass_through is already handled, so logging runs once via the canonical async path -- matching acompletion behaviour.

anneheartrecord:fix/anthropic-messages-duplicate-success-logging

3 hours ago

Add github_copilot/gpt-5.5 and github_copilot/gpt-5.4-mini (responses-only) Both models reject /v1/chat/completions with code=unsupported_api_for_model; only the /v1/responses endpoint is supported. Mirrors the existing github_copilot/gpt-5.3-codex entry.

btyu:add-copilot-gpt-5.5-and-5.4-mini

6 hours ago

Merge pull request #31140 from BerriAI/litellm_internal_staging chore(ci): promote internal staging to main

main

8 hours ago

chore(ci): bump litellm version (#31139) * bump: version 1.90.0 → 1.91.0 * adding uv lock

litellm_internal_staging

8 hours ago

fix(guardrails): content filter logs upstream errors as not_run, not guardrail failure When a provider error propagates through the content filter's streaming iterator hook, the guardrail never evaluated content. Previously the generic except arm labeled it guardrail_failed_to_respond, surfacing upstream errors as guardrail failures in logs and metrics. Catch openai.OpenAIError (the base of litellm's mapped provider exceptions), record status as not_run, and re-raise unchanged. Fixes #31004

wasimat404:fix/content-filter-not-run-on-upstream-error

10 hours ago

fix(triton): JSON-encode nested dict/list params before sending to Triton Triton's `parameters` field only accepts int/bool/string values. Extra optional params (e.g. chat_template_kwargs, used to disable model "thinking") were forwarded as raw Python objects, causing Triton to reject the request: "parameter 'chat_template_kwargs' has invalid type. It should be either 'int', 'bool', or 'string'." JSON-encode dict/list values before adding them to the parameters object so structured params survive the round trip; Triton backends can json.loads() them back. Fixes #31092

srijavuppala:fix/triton-nested-param-serialization

10 hours ago

Latest Branches

0%

fix: correct context window tokens for GPT-5 Pro and GPT-5.4 Mini/Nano#30938

2 days ago

3d0e686

xbrxr03:fix/gpt5-context-window-tokens

N/A

fix(logging): dedupe success logging for non-streaming anthropic_messages (/v1/messages)#31172

3 hours ago

e6d3d8a

anneheartrecord:fix/anthropic-messages-duplicate-success-logging

0%

Add github_copilot/gpt-5.5 and github_copilot/gpt-5.4-mini (responses-only)#31155

6 hours ago

5a213d3

btyu:add-copilot-gpt-5.5-and-5.4-mini

© 2026 CodSpeed Technology

Home Terms Privacy Docs