Claude Sonnet 4.6
Anthropic's default Sonnet model, built to approach Opus-class results at a fifth of the price
Claude Sonnet 4.6 shipped on February 17, 2026 and became the default model in claude.ai and Claude Cowork, Anthropic’s workspace product. It matches Opus 4.6’s 1 million token context window and 128,000 token output limit while pricing in at $3 per million input tokens and $15 per million output tokens, roughly a fifth of Opus 4.6’s rate. On SWE-bench Verified it scored 79.6 percent, within a few points of the Opus model despite the price gap, and it posted 59.1 percent on Terminal-Bench 2.0 and 72.5 percent on OSWorld-Verified, benchmarks that test agentic terminal and computer-use tasks. Anthropic built in extended reasoning chains that can run up to 100 steps, aimed at developers who want Opus-level reliability without paying Opus prices.