Anthropic has recently introduced Sonnet 4.6, the latest iteration of its mid-size AI model. This new version showcases enhancements in coding, instruction-following, and general computing functions, underscoring the company’s commitment to advancing AI technology.
The Sonnet 4.6 model is set to become the default choice for both Free and Pro plan users. Notable upgrades include a significantly expanded context window of 1 million tokens in the beta release, doubling the previous maximum size available. Anthropic states this expanded capacity can accommodate entire codebases, extensive contracts, or numerous research papers in a single request.
Following the Opus 4.6 launch, Anthropic plans to unveil an updated Haiku model in the near future. The Sonnet 4.6 release has already set new benchmark records, excelling in areas such as computer usage with the OS World benchmark and software engineering with the SWE-Bench benchmark. Particularly impressive is its 60.4% rating on the ARC-AGI-2 test, designed to assess skills associated with human intelligence. While Sonnet 4.6 surpasses many comparable models, it still trails behind the likes of Opus 4.6, Gemini 3 Deep Think, and a refined version of GPT 5.2.
Source: TechCrunch