Anthropic has launched Claude Sonnet 4.6, a substantial upgrade to its mid-tier AI model. This latest iteration boasts enhanced capabilities across coding, computer interaction, long-context reasoning, agent planning, knowledge work, and design. Notably, Sonnet 4.6 introduces a 1 million token context window, currently in beta, allowing for the processing of massive datasets.
For users on free and Pro plans, Claude Sonnet 4.6 is now the default model on claude.ai and Claude Cowork. Pricing remains consistent with the previous Sonnet 4.5 version. Early testers, including developers, have shown a strong preference for Sonnet 4.6 over its predecessor, with many even favoring it over Anthropic's higher-tier Claude Opus 4.5 model from late 2025. This means performance previously requiring an Opus-class model is now accessible with Sonnet 4.6, particularly for real-world office tasks.
Enhanced Coding and Computer Use
Sonnet 4.6 brings significantly improved coding proficiency to a broader user base. Developers have highlighted its consistency and superior instruction following. The model also shows marked progress in computer use skills, building on Anthropic's pioneering work in general-purpose AI interaction with software.
Anthropic's OSWorld benchmark, which simulates real-world software interaction like navigating spreadsheets and filling web forms, demonstrates Sonnet models' rapid evolution. While still not matching expert human users, the rate of improvement is notable, making AI-driven computer use more practical for diverse work tasks.
A Million Tokens of Context and Beyond
The 1M token context window in Claude Sonnet 4.6 is a game-changer, capable of holding entire codebases or extensive legal documents. Crucially, the model demonstrates effective reasoning across this vast context, aiding in complex, long-horizon planning. Its performance in evaluations like the Vending-Bench Arena, simulating business operations, showcases strategic capabilities.
Beyond benchmarks, early customers report polished visual outputs, better layouts, and improved design sensibility. Sonnet 4.6 also matches Opus 4.6 performance on OfficeQA, a benchmark for reading and reasoning over enterprise documents, marking a significant leap for document comprehension workloads. This advancement is detailed in Anthropic's announcement.
Safety and Availability
Anthropic reports that extensive safety evaluations show Sonnet 4.6 to be as safe as, or safer than, previous models, exhibiting strong safety behaviors and no major misalignment concerns. The model is available across all Claude plans, Claude Cowork, Claude Code, via API, and major cloud platforms. The free tier now includes Sonnet 4.6 with enhanced features like file creation and connectors.
