Key Highlights:
- Matches near-Opus performance at the same Sonnet pricing.
- Claude Sonnet 4.6 becomes the default model for Free and Pro users.
- Introduces a 1M token context window in beta.
- Shows major gains in coding, computer use, and long-context reasoning.
Claude Sonnet 4.6 is now live, and it marks the biggest upgrade yet to Anthropic’s Sonnet family. The new model improves coding, computer use, long-context reasoning, and agent planning, while keeping prices unchanged. For many users, performance once limited to top-tier models is now available by default.
Developed by Anthropic, Claude Sonnet 4.6 replaces Sonnet 4.5 as the default model across claude.ai and Claude Cowork. Free and Pro users now access the upgraded model automatically.
What makes Claude Sonnet 4.6 different?
Claude Sonnet 4.6 is a full capability upgrade, not a minor refresh. The model handles coding tasks more consistently. It follows instructions better. It also reasons more effectively across long documents, complex workflows, and multi-step tasks.
Anthropic says early developers prefer Sonnet 4.6 over Sonnet 4.5 by a wide margin. Many even choose it over the older Opus-class model from November 2025.
Why is the 1M token context window important?
Claude Sonnet 4.6 introduces a 1M token context window in beta. This allows users to load entire codebases, long contracts, or multiple research papers into a single prompt.
More importantly, the model maintains reasoning quality across that length. This improves long-horizon planning, document analysis, and large-scale code understanding.
How has computer use improved?
Claude Sonnet 4.6 shows major progress in computer use. It interacts with software like a human would. It clicks, types, navigates menus, and works across multiple tabs.
On the OSWorld benchmark, which tests real software use, Sonnet models have steadily improved. Early users now report near human-level performance on tasks like complex spreadsheets and multi-step web forms.
What about safety and reliability?
Anthropic says extensive safety testing shows Claude Sonnet 4.6 is as safe as, or safer than, previous models. Evaluators found strong resistance to prompt injection attacks and no major alignment concerns.
The model also produces fewer hallucinations and fewer false claims of task completion.
Pricing and availability
Claude Sonnet 4.6 keeps the same pricing as Sonnet 4.5, starting at $3 per million input tokens and $15 per million output tokens. It is available across all Claude plans, the API, and major cloud platforms.
In the conclusion, the new model signals a shift. Advanced AI capabilities are no longer limited to premium tiers. They are now mainstream.
