The latest Anthropic model brings improvements in coding, computer usage, long-context reasoning, and agent planning. Less than two weeks ago, the company released Claude Opus 4.6.
Claude Sonnet 4.6 is now the standard on the Free and Pro plans of claude.ai and Claude Cowork. For developers, the price remains $3 per million input tokens and $15 per million output tokens. Anthropic claims that users clearly prefer the new model over its predecessor.
In Claude Code, Anthropic’s development environment, early adopters preferred Sonnet 4.6 over Sonnet 4.5 about 70 percent of the time. They reported that the model reads context more effectively before modifying code and that it consolidates shared logic rather than duplicating it. Even compared to Opus 4.5, the frontier model from November 2025, users chose Sonnet 4.6 in 59 percent of cases.
Big leap in computer use
In October 2024, Anthropic was the first to introduce a general-purpose computer-using model, a type of AI that interacts with a computer’s user interface in the same way a human does. This allows the AI to understand the screen and perform tasks for the user.
Progress has been significant since the end of 2024. On OSWorld, the standard benchmark for AI computer use, Sonnet models are scoring higher and higher. This benchmark tests hundreds of tasks in real software such as Chrome, LibreOffice, and VS Code on a simulated computer.
Users now see human-level performance in tasks such as navigating complex spreadsheets or filling out web forms across multiple browser tabs. However, the model still lags behind the most skilled humans. But the progress is remarkable enough to make computer use useful for a range of work tasks.
Computer use does come with risks. Malicious actors may attempt to hijack the model by hiding instructions on websites, a technique known as prompt injection. Anthropic’s security evaluations show that Sonnet 4.6 is a significant improvement over Sonnet 4.5 and performs similarly to Opus 4.6 in resistance to prompt injections.
1 million token context window
The 1 million-token context window is large enough to cover entire codebases, long-running contracts, or dozens of research papers in a single request. More importantly, Sonnet 4.6 effectively reasons over all that context. This makes the model better at long-term planning.
This was evident in the Vending-Bench Arena evaluation, which tests how well a model can run a simulated business over time. Various AI models compete there for the highest profits. Sonnet 4.6 developed a strategy that invested heavily in capacity during the first ten simulated months and then shifted sharply toward profitability in the final sprint.
Customers reported broad improvements, with front-end code and financial analysis standing out. They described visual outputs as noticeably more polished, with better layouts, animations, and design sensibility than previous models. Fewer iteration rounds were also needed to achieve production-ready results.
Product updates and platform features
On the Claude Developer Platform, Sonnet 4.6 supports both adaptive thinking and extended thinking. Context compaction in beta automatically summarizes older context when conversations approach limits, increasing effective context length.
On the API, Claude’s web search and fetch tools now automatically execute code to filter and process search results. This ensures that only relevant content remains in the context, improving both response quality and token efficiency. In addition, code execution, memory, programmatic tool calling, tool search, and tool use examples are now available to everyone.
For Claude in Excel users, the add-in now supports MCP connectors. This allows Claude to work with tools such as S&P Global, LSEG, Daloopa, PitchBook, Moody’s, and FactSet. Users can retrieve context from outside their spreadsheet without leaving Excel. MCP connectors already set up in Claude.ai work automatically in Excel. This functionality is available on Pro, Max, Team, and Enterprise plans.
Claude Sonnet 4.6 is now available on all Claude plans, Claude Cowork, Claude Code, the API, and all major cloud platforms. The free tier has been upgraded to Sonnet 4.6 as standard and now includes file creation, connectors, skills, and compaction. Developers can get started quickly via claude-sonnet-4-6 in the Claude API.