Claude 4 Sonnet is a mid-tier large language model developed by Anthropic as part of the Claude 4 family. It is designed to provide a balance between advanced intelligence and operational speed, succeeding the Claude 3.7 iteration. The model is specialized for high-volume enterprise applications, including software engineering, document synthesis, and the orchestration of autonomous AI agents.
A defining feature of this model is its extended thinking mode, which allows it to allocate additional compute to internal reasoning processes before generating a final response. This hybrid architecture enables the model to solve more complex logical and mathematical problems by breaking them down into constituent steps. During this reasoning phase, the model can also utilize tools such as web search to verify facts and improve the accuracy of its outputs.
Claude 4 Sonnet supports a standard context window of 200,000 tokens, with expanded support for up to 1 million tokens in specific preview environments. In technical evaluations at release, the model demonstrated high performance in coding tasks, achieving a 72.7% score on the SWE-bench Verified benchmark. It also features enhanced steerability and more precise instruction following compared to previous Sonnet versions.