Anthropic has introduced Sonnet 4.6, an updated version of its midsized Sonnet model, aligning with the company’s four-month update cycle. This release brings significant advancements in various capabilities.
The company highlighted that Sonnet 4.6 features improvements in coding, instruction-following, and computer use capabilities. This model will serve as the default for users on Anthropic’s Free and Pro plans, indicating its enhanced usability and performance.
Sonnet 4.6, currently in its beta release, boasts a 1 million-token context window, doubling the largest previously available capacity for Sonnet. Anthropic stated that this expanded window can accommodate “entire codebases, lengthy contracts, or dozens of research papers in a single request,” thereby enhancing the model’s versatility and application scope.
The launch of Sonnet 4.6 follows the introduction of Opus 4.6 two weeks prior, suggesting a rapid progression in Anthropic’s model updates. Additionally, an updated Haiku model is anticipated in the near future, indicating a consistent development trajectory.
Sonnet 4.6 has achieved new benchmark scores in several key areas, including OS World for computer use and SWE-Bench for software engineering. Notably, it scored 60.4% on ARC-AGI-2, a measure of human-specific intelligence skills. This score positions Sonnet 4.6 above most comparable models, although it ranks below Opus 4.6, Gemini 3 Deep Think, and a refined variant of GPT 5.2.
