Claude Sonnet 4

Claude Sonnet 4 is a hybrid reasoning large language model from Anthropic, introduced in the May 2025 Claude 4 system card alongside Claude Opus 4. The system card describes Sonnet 4 as supporting an extended thinking mode that lets developers control how much deliberate reasoning the model invests in a response, and it was deployed under Anthropic's AI Safety Level 2 Standard. Its design targets work that benefits from careful step-by-step reasoning paired with broad context, including coding, tool use, and analysis of large document collections, making it a practical middle-tier option in the Claude 4 family for production assistants and research-style tasks that need more deliberation than a non-reasoning model.

After its initial release, Anthropic expanded Claude Sonnet 4's capabilities with a 1 million token context window, a 5x increase announced in August 2025 and made available in public beta on the Anthropic API, with subsequent availability on Amazon Bedrock and Google Cloud's Vertex AI. The longer window is aimed at workloads such as whole-codebase analysis spanning tens of thousands of lines, multi-document synthesis across research papers or legal contracts, and agents that need to stay coherent across hundreds of tool calls. The system card also documents standard training characteristics like training data and process notes, crowd worker practices, carbon footprint reporting, and usage policy framing, giving teams a published reference for how the model was prepared and governed.