The first hybrid reasoning model let you see Claude's thinking before it answered — on every tier except the free one.
Being able to read the reasoning before you act on it is what makes an AI-drafted deal assessment auditable rather than something you have to take on trust.
Anthropic released Claude 3.7 Sonnet on 24 February 2025, describing it as the first hybrid reasoning model on the market, with visible extended thinking. It shipped across Free, Pro, Team and Enterprise, plus the Claude Developer Platform, Amazon Bedrock and Google Cloud Vertex AI. Pricing held at $3 per million input tokens and $15 per million output tokens, thinking tokens included.
Extended thinking is available on every surface except the free tier. Claude Code was introduced alongside it as a limited research preview.
Two minutes, once a week. What changed in AI, and what to run because of it.