Claude Sonnet 4.6: The AI That Can Finally Use Your Computer

Anthropic has officially released Claude Sonnet 4.6, a massive midcycle upgrade that pushes the boundaries of what a mid tier AI model can do. While it carries the Sonnet name, this version is a full scale overhaul that targets the hardest problems in AI, from complex coding and agent planning to long context reasoning. It has now become the default model for Free and Pro users on Claude.ai and Claude Cowork, maintaining the same price point as the previous 4.5 version while offering performance that often rivals or exceeds the flagship Opus model.

What makes Sonnet 4.6 stand out is its specialized focus on computer use, a capability that allows the AI to interact with software just like a human would. Rather than relying on rigid APIs, the model can view a virtual screen, move a mouse, and type on a keyboard to navigate apps like Chrome, VS Code, and LibreOffice. While Anthropic acknowledges the tech is still evolving, early tests on the OSWorld benchmark show the model can handle multistep web forms and complex spreadsheets with human level proficiency. This shift matters because it allows AI to automate specialized, legacy software that was never built for modern automation.

On the coding front, Sonnet 4.6 is proving to be a favorite among developers. Early testing shows that users prefer it over Sonnet 4.5 roughly 70% of the time, noting that it is much better at understanding context before suggesting code changes and consolidating shared logic rather than duplicating it. Surprisingly, it even beat out the high end Opus 4.5 model 59% of the time, with users reporting that Sonnet 4.6 is less prone to “laziness” and far more consistent when following complex, multistep instructions. It’s a rare case where the more efficient model is also becoming the more reliable one for daily work.

For those dealing with massive amounts of data, the new 1 million token context window—currently in beta—is a game changer. This allows the model to digest entire codebases, dozens of research papers, or lengthy legal contracts in a single go. In high stakes business simulations like the Vending Bench Arena, Sonnet 4.6 demonstrated sophisticated long horizon planning by aggressively investing in growth before pivoting to profitability to beat the competition. It isn’t just processing text; it’s executing strategy.

Anthropic is also bringing these tools into the apps where people already work. A new Excel add in for Pro, Max, Team, and Enterprise users now supports MCP connectors, allowing Claude to pull data from external sources like Moody’s, PitchBook, and S&P Global without ever leaving the spreadsheet. While computer use does introduce new risks like prompt injections, Anthropic has bolstered the model’s safety protocols, making Sonnet 4.6 as resilient as its top tier Opus 4.6 counterpart. It’s a powerful, polished update that makes high end AI feel more practical for everyday tasks.