Anthropic has released a new version of its midsize Sonnet model to coincide with the company’s four-month update cycle. In a post announcing the new model, Anthropic highlighted improvements in how it codes, follows instructions and uses computers.
Sonnet 4.6 will be the default model for users on the Free and Pro plans.
The beta release of Sonnet 4.6 includes a 1 million token context window, which is twice the size of the largest window previously available in Sonnet. Anthropic described the new context window as “sufficient to hold an entire codebase, long contract, or dozens of research papers in a single request.”
This release comes just two weeks after the release of Opus 4.6, and updated Haiku models are likely to follow in the coming weeks.
The announcement comes with a series of new record benchmark scores, including OS World for Computing and SWE-Bench for Software Engineering. But perhaps most impressive is its 60.4% score on the ARC-AGI-2, which aims to measure skills specific to human intelligence. On this score, the Sonnet 4.6 outperforms most comparable models, but still falls behind models like the Opus 4.6, Gemini 3 Deep Think, and one sophisticated version of GPT 5.2.
Source link
