Our most intelligent model to date
Anthropic’s AI coding assistant, designed for deep context understanding and capable of handling complex software tasks with a massive context window (up to 200K tokens).
Claude 3.7 Sonnet makes 45% fewer unnecessary refusals compared to predecessor by making more nuanced distinctions between harmful and benign requests. Improved safety evaluation and resistance to prompt injection attacks documented in system card, enabling more reliable production deployments.
Claude Code provides command-line tool for agentic coding with ability to search/read code, edit files, write/run tests, commit/push to GitHub, and use command-line tools. GitHub integration available on all Claude plans enables developers to connect repositories directly for bug fixing, feature development, and documentation across projects.
Claude 3.7 Sonnet available on all Claude plans (Free, Pro, Team, Enterprise), Claude Developer Platform, Amazon Bedrock, and Google Cloud Vertex AI. Provides flexibility for organizations with existing cloud commitments or multi-cloud strategies.
Claude 3.7 Sonnet achieves state-of-the-art performance on SWE-bench Verified for real-world software issues. Claude Code enables developers to delegate substantial engineering tasks directly from terminal, completing tasks in single pass that normally take 45+ minutes. Both tools excel at handling complex codebases, full-stack updates, and production-ready code generation.
Claude 3.7 Sonnet offers both standard mode for near-instant responses and extended thinking mode for step-by-step reasoning visible to users. API users control thinking budget (up to 128K tokens), allowing trade-off between speed, cost, and answer quality based on task requirements.