Anthropic has announced the release of Claude Sonnet 4.6, marking a significant advancement in AI model capabilities. This latest version offers enhanced coding skills, improved computer automation, and advanced reasoning capabilities. With a score of 72.5% on the OSWorld-Verified benchmark—an increase from the previous 28.0—it stands out in both financial analysis and office tasks, surpassing the expensive Opus 4.6 model.
Claude Sonnet 4.6 is now available across various platforms, including GitHub Copilot, Windsurf, Microsoft Foundry, Amazon Bedrock, and Kiro IDE and CLI. It’s designed for integration with platforms like Microsoft Foundry to support enterprise governance and compliance infrastructure, offering functionalities such as browser automation and legacy system integration. This model is also part of GitHub Copilot, available to Copilot Pro, Pro+, Business, and Enterprise users with appropriate policy settings in place.
Key improvements in the Claude Sonnet 4.6 model include a 1 million token context window (available in beta), improved prompt injection resistance, and enhanced safety evaluations showing the model’s prosocial behavior. It’s been optimized for better instruction following, consistency, and multi-step task execution. The model manages complex tasks like navigating spreadsheets, filling web forms, managing codebases, and executing long-horizon reasoning.
Besides its technical upgrades, Claude Sonnet 4.6 is rolling out with cost-efficient pricing akin to Sonnet 4.5, but with Opus-level performance. It is available with promotional multipliers in environments such as Windsurf and requires only minor prompt adjustments when migrating from Sonnet 4.5.
Early users have demonstrated a preference for Sonnet 4.6 in coding tasks, with a notable 70% preferring it over the previous Claude Sonnet 4.5 version, and 59% favoring it over Opus 4.5.
Overall, Claude Sonnet 4.6 sets a new standard for mid-tier models, combining affordability with high-performance features and extensive integration capabilities.

Collections

Anthropic has released Claude Sonnet 4.6, a mid-tier AI model featuring enhanced coding capabilities, computer automation, and advanced reasoning. It achieves a 72.5% score on the OSWorld-Verified benchmark, up from 28.0, outperforming the more expensive Opus 4.6 in financial analysis and office tasks. Key features include a 1 million token context window (beta), improved prompt injection resistance, and better multi-step task execution. The model is available across GitHub Copilot, Windsurf, Microsoft Foundry, Amazon Bedrock, and Kiro IDE at pricing comparable to Sonnet 4.5. Early user preference data shows 70% favor it over Sonnet 4.5 for coding tasks and 59% prefer it over Opus 4.5.

Comprehensive Overview of Claude Sonnet 4.6