Anthropic releases Sonnet 4.6

Anthropic has officially launched the latest iteration of its mid-sized Sonnet model, marking a significant update in line with the company's commitment to releasing new versions every four months. In their announcement, Anthropic highlighted several key advancements in coding capabilities, adherence to instructions, and overall computer functionality. The newly introduced Sonnet 4.6 will be the standard model for both Free and Pro plan users. Notably, the beta version boasts an impressive context window of 1 million tokens—double the size of the previously available maximum. This expanded context capacity is described by Anthropic as capable of accommodating entire codebases, lengthy legal contracts, or multiple research papers in a single request. This release follows closely on the heels of Opus 4.6, which was launched just two weeks prior, and signals that an updated Haiku model is likely to be unveiled soon. Accompanying this new Sonnet version are record-breaking benchmark scores, including top marks in OS World for computer usage and SWE-Bench for software engineering tasks. Among its notable achievements, Sonnet 4.6 has achieved an impressive score of 60.4% on the ARC-AGI-2 benchmark, which assesses skills closely aligned with human intelligence. While Sonnet 4.6 surpasses many of its competitors in several aspects, it still falls short of leading models such as Opus 4.6, Gemini 3 Deep Think, and a refined edition of GPT 5.2. Nonetheless, its advancements position it as a formidable player in the evolving landscape of AI technologies.