You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Call it through the LiteLLM AI Gateway with the request you already send, on Anthropic, Bedrock, Gemini Enterprise Agent Platform and Azure, with spend tracking, rate limits, fallbacks and logging in one place.
20% cheaper than Opus 5 - $4 / MTok input and $20 output, against $5 and $25, with the 1M context window and 128K max output unchanged. Batch is $2 and $10.
Cache reads - $0.20 / MTok. Cache writes are $5 for five minutes and $8 for an hour.
Worth checking before you route traffic to it:
Forced tool use is rejected. The cost map row carries supports_forced_tool_use: false, so with drop_params on LiteLLM downgrades tool_choiceany and tool to auto and logs a warning, and without it you get a clean client-side 400. The downgrade succeeds but leaves the model free not to call your tool, so move to auto with strict tool use or structured outputs.
Thinking cannot be disabled. thinking: {"type": "disabled"} and manual budget_tokens both 400. Effort is the only control, and its default is medium where Opus 5's was high, so a request that never set effort now thinks less.
Thinking blocks stay tied to the model. On the Claude API only Fable 5.1 and Mythos 5.1 read Opus 5.5's blocks, so check which direction your fallback chain runs.
On the Claude API and Gemini Enterprise Agent Platform, the older computer_20251124 tool is rejected; declare computer_toolset_20260801 instead. On Bedrock the old tool still works.
Pricing landed in PR #42489 for the first-party route. The Bedrock, Gemini Enterprise Agent Platform and Azure ids run but are not priced yet, so spend on those logs as $0 until their rows land. Reload your pricing in the UI under Models + Endpoints -> Price Data -> Reload Price Data (or POST /reload/model_cost_map as an admin), on any version v1.76.0 or newer.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
Claude Opus 5.5 is now supported on LiteLLM
Call it through the LiteLLM AI Gateway with the request you already send, on Anthropic, Bedrock, Gemini Enterprise Agent Platform and Azure, with spend tracking, rate limits, fallbacks and logging in one place.
20% cheaper than Opus 5 - $4 / MTok input and $20 output, against $5 and $25, with the 1M context window and 128K max output unchanged. Batch is $2 and $10.
Cache reads - $0.20 / MTok. Cache writes are $5 for five minutes and $8 for an hour.
Worth checking before you route traffic to it:
Forced tool use is rejected. The cost map row carries
supports_forced_tool_use: false, so withdrop_paramson LiteLLM downgradestool_choiceanyandtooltoautoand logs a warning, and without it you get a clean client-side 400. The downgrade succeeds but leaves the model free not to call your tool, so move toautowith strict tool use or structured outputs.Thinking cannot be disabled.
thinking: {"type": "disabled"}and manualbudget_tokensboth 400. Effort is the only control, and its default ismediumwhere Opus 5's washigh, so a request that never set effort now thinks less.Thinking blocks stay tied to the model. On the Claude API only Fable 5.1 and Mythos 5.1 read Opus 5.5's blocks, so check which direction your fallback chain runs.
On the Claude API and Gemini Enterprise Agent Platform, the older
computer_20251124tool is rejected; declarecomputer_toolset_20260801instead. On Bedrock the old tool still works.Get started:
Pricing landed in PR #42489 for the first-party route. The Bedrock, Gemini Enterprise Agent Platform and Azure ids run but are not priced yet, so spend on those logs as $0 until their rows land. Reload your pricing in the UI under Models + Endpoints -> Price Data -> Reload Price Data (or POST /reload/model_cost_map as an admin), on any version v1.76.0 or newer.
Read the full guide → Day 0 Support: Claude Opus 5.5
Misbah Syed
DevRel @ LiteLLM
All reactions