Anthropic has launched Claude Sonnet 4.6, its most capable Sonnet model yet, with a 1M-token context window in beta. The company announced the model on 8 November 2026, describing it as a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work and design.

The headline change is the size of the context window. A million tokens can hold an entire codebase, a lengthy contract or dozens of research papers in a single request. Anthropic says the important point is not only the size, but that Sonnet 4.6 can reason effectively across that much context. A large window is of little use if a model loses the thread part-way through a long document or project.

Coding is where Anthropic makes its strongest claims. In early testing inside Claude Code, users preferred Sonnet 4.6 over Sonnet 4.5 roughly 70 percent of the time, according to the company. Users also preferred it to Opus 4.5, the frontier model from November 2025, 59 percent of the time. Anthropic says testers found the model read the context before changing code, consolidated shared logic instead of duplicating it, and was less prone to overengineering, with better instruction following and more consistent follow-through on multi-step tasks.

Those preference figures come from Anthropic's own early testing with its own users, not from independent benchmarks. They should be treated as the company's claims until independent testing and real-world reports confirm how the model behaves in other people's codebases. The same caution applies to the benchmark gains Anthropic reports for computer use.

On computer use, Anthropic reports a major step forward on OSWorld, the benchmark in which a model performs tasks across real software such as Chrome, LibreOffice and VS Code by clicking and typing like a person. The company says early users are seeing human-level capability on tasks such as navigating a complex spreadsheet or filling out a multi-step web form, then pulling the work together across several browser tabs. It also says the model still lags behind the most skilled humans, and that prompt-injection resistance is a major improvement over Sonnet 4.5 and now performs similarly to Opus 4.6.

Pricing is unchanged from Sonnet 4.5, starting at $3 per million input tokens and $15 per million output tokens. Sonnet 4.6 is now the default model in claude.ai and Claude Cowork on the Free and Pro plans. It is available now on all Claude plans, Claude Cowork, Claude Code, the API under the model ID claude-sonnet-4-6, and all major cloud platforms.

The free tier also gains file creation, connectors, skills and compaction alongside the new default model. Context compaction, in beta, automatically summarises older context as conversations approach their limits, increasing the effective context length. On the Claude Platform, Sonnet 4.6 supports adaptive thinking and extended thinking, while the API's web search and fetch tools now automatically write and execute code to filter and process search results. The Claude add-in for Excel now supports MCP connectors, including S&P Global, LSEG, Daloopa, PitchBook, Moody's and FactSet, on Pro, Max, Team and Enterprise plans.

Anthropic is clear about when to reach for its larger model instead. Opus 4.6 remains the strongest option for the deepest reasoning tasks, such as large codebase refactoring, coordinating multiple agents in a workflow and problems where getting the answer exactly right matters most. The practical message of this launch is that Sonnet 4.6 should now handle most daily work, with Opus kept in reserve for the hardest problems.

The 1M-token window is in beta, so rough edges should be expected, and more context also costs more per request, since every token sent is a token paid for. For developers who write code for a living, the launch still changes the calculation: the mid-tier model now does work that previously needed the most expensive model in the line-up, at the same price as its predecessor.