Anthropic has released Sonnet 4.6, the latest version of its midsized artificial intelligence model, continuing the company’s four-month model update cycle. The new model introduces improvements in coding performance, instruction-following, and computer-use capabilities, and will become the default model for Free and Pro users.
The beta release expands the model’s context window to 1 million tokens, enabling analysis of large codebases, contracts, and research materials within a single request. The launch follows the recent release of Opus 4.6, with an updated Haikumodel expected soon. Anthropic reported strong benchmark results for Sonnet 4.6, including performance gains in software engineering, computer-use evaluations, and reasoning-focused AI intelligence tests.
