Claude 4 Sonnet is a multimodal large language model developed by Anthropic, released on May 22, 2025. It serves as the mid-tier performance model in the Claude 4 family, succeeding Claude 3.7 Sonnet. The "non-reasoning" designation distinguishes this model's standard inference mode—which provides direct, low-latency responses—from the "extended thinking" mode that uses a dedicated reasoning budget for complex deliberation.
The model features a 1 million token context window and is optimized for agentic workflows, software engineering, and high-volume data processing. It natively supports the Model Context Protocol (MCP), enabling it to autonomously interact with external tools, local files, and APIs. Upon release, Claude 4 Sonnet demonstrated significant technical proficiency, achieving a 72.7% success rate on the SWE-bench for coding tasks.
Developed using Constitutional AI and adhering to AI Safety Level 3 (ASL 3) protocols, Claude 4 Sonnet is designed for reliable and steerable enterprise applications. It supports multimodal inputs, including text and image processing, and is positioned for production environments where a balance of frontier-level intelligence and operational efficiency is required.