Anthropic on February 17 launched Claude Sonnet 4.6, calling it its most capable Sonnet model to date, with upgrades across coding, computer use, long-context reasoning, agent planning, knowledge work and design.
Claude Sonnet 4.6 is available across Claude plans, Claude Cowork, Claude Code, the company’s API and major cloud platforms. Developers can access it through the Claude API using the model name ‘claude-sonnet-4-6’.
The company said Sonnet 4.6 is now the default model for users on Free and Pro plans in Claude.ai and Claude Cowork. Pricing remains unchanged from Sonnet 4.5, starting at $3 per million input tokens and $15 per million output tokens.
Sonnet 4.6 includes a one-million-token context window in beta, allowing users to input entire codebases, contracts or multiple research papers in a single request. The company said the model can reason across that context and support longer planning tasks.
“Sonnet 4.6 brings much-improved coding skills to more of our users. Improvements in consistency, instruction following, and more have made developers with early access prefer Sonnet 4.6 to its predecessor by a wide margin,” the company said in a blog post.
In early testing within Claude Code, users preferred Sonnet 4.6 over Sonnet 4.5 about 70% of the time, according to the company. Users also preferred it over Claude Opus 4.5 in 59% of cases. They reported fewer hallucinations, fewer false claims of task completion and better instruction following.
Anthropic said the model shows gains in computer use, an area it first introduced in October 2024. Computer use allows the model to interact with software interfaces by clicking and typing on a simulated computer, instead of relying on APIs.
On OSWorld, a benchmark that tests AI interaction with real software such as Chrome, LibreOffice and VS Code, Sonnet models have shown steady improvements over the past sixteen months.
The company said early users of Sonnet 4.6 have demonstrated human-level capability in tasks such as navigating spreadsheets and completing multi-step web forms. It added that the model still lags behind highly skilled human users.
Anthropic also flagged risks tied to computer use, including prompt injection attacks where hidden instructions on websites attempt to manipulate model behaviour. The company said safety evaluations show Sonnet 4.6 is more resistant to such attacks than Sonnet 4.5 and performs similarly to Opus 4.6.
“Our safety researchers concluded that Sonnet 4.6 has ‘a broadly warm, honest, prosocial, and at times funny character, very strong safety behaviours, and no signs of major concerns around high-stakes forms of misalignment,’” the company said.
New features
Sonnet 4.6 supports adaptive thinking, extended thinking and context compaction in beta on the Claude Developer Platform. Context compaction summarises older conversation history as token limits are reached.
On the API, Anthropic said Claude’s web search and fetch tools can now automatically write and execute code to filter and process search results. Code execution, memory, programmatic tool calling, tool search and tool use examples are now generally available.
For Excel users, Claude’s add-in now supports MCP connectors, enabling access to external data sources such as S&P Global, LSEG, Daloopa, PitchBook, Moody’s and FactSet within spreadsheets. This feature is available on Pro, Max, Team and Enterprise plans.
Your reaction
Discussion
What's your take on this story?
No comments yet — be the first to share your view.