Claude Sonnet 5.5 is out at Sonnet 5 prices. Code written for Sonnet 5 can break on it in five documented ways, so change the model string only after checking them.
Anthropic launched claude-sonnet-5-5 on 28 September on its API, Bedrock, Google Cloud and Microsoft Foundry.1 List price is unchanged from Sonnet 5: $2 input, $10 output and $0.20 cache reads per million tokens. Anthropic says it costs up to 30% less per task because it uses fewer tokens.2 That per-task figure is Anthropic's own measurement.
The release notes list five breaks.1 thinking: {"type": "disabled"} must become {"type": "between_tools"}. Forced tool use (tool_choice any or tool) returns a 400. On the Claude API and Google Cloud the older computer_20251124 tool is refused. The advisor tool rejects Opus 4.8, Opus 4.7 and Sonnet 5 as advisors. Thinking blocks are tied to the model, the conversation and the account that produced them.
Nothing fails until you switch models. Before switching, search your code for "disabled" thinking and forced tool_choice, and test any flow that moves a conversation between accounts.
The account-binding change, and what GitHub measured
Thinking blocks from Sonnet 5.5 work only in the account that produced them or a linked one. When another account sends one, the API drops the block before the model sees it and the request still succeeds.1 Nothing errors, so a pipeline that hands conversations between accounts loses reasoning context with no signal. Anthropic's launch post names switching accounts mid-session in Claude Code as one case.2
GitHub made the model generally available in Copilot the same day, billed at provider list price, with a gradual rollout.3 Its own description is more modest than Anthropic's: it says Sonnet 5.5 matched Sonnet 5 on coding tasks while using significantly fewer steps, tokens and tool calls.3