Anthropic launches Claude Sonnet 4.6 with coding, reasoning upgrades
What's the story
Anthropic has unveiled the latest version of its mid-size AI model, Sonnet 4.6. The upgrade comes just four months after the last update and focuses on enhancing coding, instruction-following capabilities, and computer usage. The new model will be the default for both Free and Pro plan users on Anthropic's platform.
Enhanced capabilities
New model boasts 1 million token context window
The beta release of Sonnet 4.6 comes with a whopping one million token context window, double the size of the previous largest window for Sonnet. This new feature is capable of accommodating entire codebases, long contracts, or multiple research papers in a single request. The model also boasts record benchmark scores on OS World for computer use and SWE-Bench for software engineering tasks.
Competitive edge
Sonnet 4.6 outperforms most rival models
Sonnet 4.6 has scored an impressive 60.4% on the ARC-AGI-2 benchmark, which measures human intelligence-specific skills. This puts it ahead of most comparable models in the market. However, it still lags behind top-tier models like Opus 4.6, Gemini 3 Deep Think, and a refined version of GPT 5.2 Pro in this particular category.