Claude Sonnet 4.6 Doubles Context Window, Sets Coding Benchmarks
Anthropic's Claude Sonnet 4.6 now features a 1 million token context window and sets new records on key coding benchmarks.Summary
Anthropic released Claude Sonnet 4.6 on February 17, 2026, making it the default model for Free and Pro plans on claude.ai. This mid-tier model doubles its context window to 1 million tokens (currently in beta), allowing it to process massive inputs like entire codebases, while maintaining the same pricing as its predecessor ($3/M input, $15/M output). Internally, users preferred Sonnet 4.6 over the previous flagship, Opus 4.5, 59% of the time. The model shows significant coding improvements, including reduced duplication and hallucinations, and sets record scores on the SWE-Bench and OSWorld benchmarks, although it trails higher-tier models like Opus 4.6 on the ARC-AGI-2 reasoning benchmark. The maximum output token limit has been increased to 64K.
(Source：onllm.dev)