Claude 3.7 Sonnet & Claude Code Released

Anthropic has launched Claude 3.7 Sonnet, described as its most intelligent model yet and the market's first "hybrid reasoning" AI. This model uniquely integrates both near-instant responses and an optional "extended thinking" mode where its step-by-step reasoning process is made visible, enhancing performance on complex tasks like math, coding, and instruction-following. API users gain fine-grained control over this thinking process via a token budget.

Claude 3.7 Sonnet shows significant improvements, particularly in coding and front-end development, achieving state-of-the-art results on benchmarks like SWE-bench (real-world software issues) and TAU-bench (complex agent tasks). Anthropic emphasizes a focus on practical, real-world business use cases rather than just competition problems.

Alongside the model, Anthropic introduced Claude Code, a limited research preview of an agentic coding tool for the command line. Claude Code aims to automate substantial engineering tasks by searching codebases, editing files, running tests, managing commits, and using CLI tools autonomously, while keeping the developer informed.

Claude 3.7 Sonnet is available across all Claude plans (Free tier excluded from extended thinking), the API, Amazon Bedrock, and Google Cloud Vertex AI, maintaining the previous Sonnet pricing ($3/M input, $15/M output). The release also brings improved safety, reducing refusals by 45%, and makes GitHub integration available on all Claude.ai plans.