A Full Upgrade Across Core Capabilities
Anthropic has launched Claude Sonnet 4.6, the latest iteration of its mid-tier model that balances capability with cost-efficiency. The model represents a significant jump in performance across coding, computer use, long-context reasoning, agent planning, knowledge work, and design tasks. It's now the default model for Free and Pro plan users on claude.ai and Claude Cowork.
Coding and Instruction Following
Early testing shows developers strongly prefer Sonnet 4.6 over Sonnet 4.5 (70% of the time) and even over the frontier Opus 4.5 model (59% of the time). Users report that Sonnet 4.6:
- More effectively reads context before modifying code
- Consolidates shared logic rather than duplicating it
- Shows significantly less overengineering and "laziness"
- Is substantially better at instruction following
- Produces fewer false claims of success and hallucinations
- Delivers more consistent follow-through on multi-step tasks
This performance-at-cost profile means organizations can now accomplish work previously requiring expensive Opus-class models using the more economical Sonnet tier.
Computer Use and Extended Context
Sonnet 4.6 marks a major improvement in computer use capabilities, advancing on the OSWorld benchmark from 20.1% (Sonnet 4.5) to higher performance levels. The model can now handle human-level tasks like navigating complex spreadsheets and filling out multi-step web forms before coordinating across multiple browser tabs—all without special APIs or connectors.
The 1M token context window (in beta) is a practical upgrade, allowing the model to reason effectively across entire codebases, lengthy contracts, or dozens of research papers in a single request. Early testing via the Vending-Bench Arena evaluation demonstrated Sonnet 4.6's sophisticated long-horizon planning: the model invested heavily in capacity early on, then pivoted to profitability to outcompete other AI models.
Safety and Pricing
Anthropic has completed extensive safety evaluations showing Sonnet 4.6 is as safe as or safer than recent Claude models. The model demonstrates strong resistance to prompt injection attacks compared to Sonnet 4.5 and performs similarly to Opus 4.6. Pricing remains unchanged from Sonnet 4.5: $3 per million input tokens and $15 per million output tokens on the API.
Developers can start using Sonnet 4.6 immediately on claude.ai and via the API at the standard Sonnet pricing tier.