Claude 3.7 Sonnet
The first hybrid reasoning model: instant answers or extended thinking in one model
About
Claude 3.7 Sonnet, launched February 24, 2025, was Anthropic's answer to the model-picker problem: instead of separate fast and thinking models, one model does both. Standard mode answers quickly; extended thinking mode shows a visible scratchpad and lets API users set a thinking budget to trade latency and cost for depth. It landed as the coding model to beat, with strong SWE-bench and TAU-bench results at launch, and it arrived alongside the Claude Code agentic coding tool preview. Extended thinking was reserved for paid plans. Pricing stayed flat at $3/$15 per million tokens, which made the hybrid design feel like a free upgrade over 3.5 Sonnet.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Team-reported: 62.3% SWE-bench Verified (70.3% with high-compute scaffolding); state-of-the-art on TAU-bench at launch
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: Launch date Feb 24, 2025 and hybrid-reasoning design corroborated across 5+ sources; Pricing $3/$15 and 200K context consistent everywhere; SWE-bench 62.3% (70.3% scaffolded) and TAU-bench SOTA reported at launch, labeled team-reported; Free-tier limitation (no extended thinking on free) captured honestly.