Claude Sonnet 5.5 keeps Sonnet 5's price, and its cache reads were halved on 7 October
Claude Sonnet 5.5 was released on 28 September 2026[1,2]. The model id is claude-sonnet-5-5[1,2].
In Anthropic's description, Opus 5.5 is for complex work that needs careful judgment, while Sonnet 5.5 is strongest at well-scoped everyday tasks, fixing bugs, and creating documents, slides and spreadsheets[1].
Per million tokens it costs $2 input and $10 output, unchanged from Sonnet 5[1]. Cache writes cost $2.50, and cache reads were $0.20 at launch, the same as Sonnet 5[1].
On 7 October cache reads dropped to $0.10, going from 10 percent of the input price to 5 percent, the same ratio as Opus 5.5[1,2]. Because cache reads make up a large share of token consumption, Anthropic estimates that the cut makes Sonnet 5.5 about 20 percent cheaper on most agentic tasks[3].
Sonnet 5.5 often finishes the same work with far fewer tokens than Sonnet 5[1]. In Anthropic's testing, the cost per task fell by up to 30 percent[1]. Output is also generated more than 30 percent faster than on Sonnet 5[1].
Terminal-Bench 4.0 measures whether a model can complete multi-step professional tasks in a command-line interface[1]. Sonnet 5.5 scores 70.6 percent, up from 10.3 percent for Sonnet 5 and above the 66.4 percent of Opus 5.5 at xhigh effort[1].
Anthropic adds that benchmark scores capture only one side of a model, and that Opus 5.5 remains clearly stronger at complex work requiring sustained judgment[1].
The context window is 1 million tokens and one response can be up to 128K tokens, both the same as Opus 5.5[4]. The training data and knowledge cutoffs are both June 2026, and retirement will not come before 28 September 2027[4].
The default effort is high on the Claude Platform and medium in Claude Code and the Claude apps[1].
Some code written for Sonnet 5 will not work unchanged on Sonnet 5.5[2]. To turn off up-front thinking at high effort or below, set the thinking type to between_tools instead of disabled[1,2].
Forcing tool use with the tool_choice types any and tool returns a 400 error[2]. On the Claude API and Google Cloud, the earlier computer use tool computer_20251124 is not accepted[2].
Thinking blocks from Sonnet 5.5 work only in the account that produced them or in an account linked to it[1,2]. If another account sends one, the API drops the block and the request still succeeds[2].
On 30 September Anthropic announced the deprecation of Claude Sonnet 4.5 (claude-sonnet-4-5-20250929) and recommended moving to Sonnet 5.5[2]. Sonnet 4.5 is scheduled to retire from the Claude API on 30 November 2026[2].