AnthropicClaude Sonnet

Claude Sonnet 5.5

Claude Sonnet 5.5 targets everyday coding, writing and agent work. Existing applications need to check their request options: familiar sampling settings are not accepted by this model.

The lowest thinking mode still has a specific purpose

Adaptive thinking is enabled by default. The between_tools setting removes up-front thinking and works at high effort or below. It should not be treated as the same control as temperature.Claude Sonnet 5.5 reference

Review the request builder as well as the prompt

Practical guidance. When migrating an application, inspect shared request defaults. An unused temperature field can still cause an invalid request. Keep model-specific capability checks close to the code that builds the request.Claude Sonnet 5.5 reference

Use a revision task with observable requirements

Practical guidance. For a product description, provide the product facts, intended reader and statements that need evidence. Ask for a revision that preserves those facts. Judge whether the result answers the reader’s question, rather than whether it sounds polished.

Specifications & access

Claude Sonnet 5.5 specifications
DetailValue and conditions
Launch date
AccessAPI
Model identifiersclaude-sonnet-5-5
Context window1,000,000 tokens

Context limit

Claude Sonnet 5.5 reference
Maximum output128,000 tokens

Ordinary maximum output

Claude Sonnet 5.5 reference
Input price2 USD per million tokens

Standard input rate

Claude Sonnet 5.5 reference
Output price10 USD per million tokens

Standard output rate

Claude Sonnet 5.5 reference
Unsupported parameterstemperature, top_p and top_k return a 400 error.Claude Sonnet 5.5 reference
Default efforthighClaude Sonnet 5.5 reference
Minimum cacheable prompt512 tokensClaude Sonnet 5.5 reference

Related models

  • A model for extended coding and knowledge work with vision and computer-use support.

    Launch date API
  • A model for high-volume classification, extraction and routing tasks.

    Launch date API