Claude Unveils Sonnet 4.6: A Game Changer in Performance and Accessibility!

Anthropic is continually advancing its range of models. Following the release of Opus 4.6 two weeks ago, the company is now unveiling Claude Sonnet 4.6, modestly described as “its most efficient Sonnet model to date.”

A Sonnet Model That Matches Opus, Without Extra Costs

Claude Sonnet 4.6 sets clear goals: to match the performance of Opus 4.5 while maintaining the price point of the Sonnet line. This model features a comprehensive upgrade in terms of programming, reasoning, and knowledge processing. Sonnet 4.6 also incorporates a context window of one million tokens in beta, which is double that of its predecessor. In practical terms, tasks that previously required Opus, such as analyzing complex documents, refactoring code, or multi-step planning, are now available at the cost of a Sonnet, namely 3 dollars per million tokens for input and 15 dollars per million tokens for output.

Developers who had early access significantly prefer Sonnet 4.6 over its predecessor. They often favor it even over our smartest model from November 2025, Claude Opus 4.5., Anthropic states.

As a result, Sonnet 4.6 is now the default model for both free and Pro users, whether through Claude apps, Cowork, Claude Code, or the API.

Computer Use: An AI That Operates a Computer Like a Human

With Sonnet 4.6, Anthropic has made significant strides in the model’s ability to autonomously use a computer. In the OSWorld benchmark, which assesses an AI’s capability to perform tasks on a computer, Sonnet 4.6 has achieved 72.5%, compared to 14.9% when this feature was introduced in October 2024.

Indeed, the model can navigate through complex spreadsheets, fill out multi-step web forms, or switch between different tabs without any technical connectors needed. It clicks and types just as a user would.

Early users of Sonnet 4.6 report nearly human-like performance in tasks like navigating a complex spreadsheet or filling out a multi-step web form, concludes Anthropic.