Anthropic has released Claude Sonnet 4.6, its latest midsized model, with improvements to coding, instruction-following and computer use. The update doubles Sonnet's context window to one million tokens and makes it the default for Free and Pro plan users. Sonnet 4.6 posted record benchmark scores on OS World and SWE-Bench, and achieved 60.4% on ARC-AGI-2, placing it above most comparable models but still behind Opus 4.6, Gemini 3 Deep Think and a refined version of GPT 5.2. The release follows Opus 4.6's launch two weeks earlier.
Anthropic releases Sonnet 4.6 - TechCrunch