Claude Sonnet 4.6

2026-02-17 · Publisher : Anthropic · Domain : Artificial Intelligence

Anthropic launches Claude Sonnet 4.6, its most capable Sonnet, improved in coding, computer use, long-context reasoning and agent planning, at the same price as Sonnet 4.5.

What changes

  • Full upgrade: coding, computer use, long-context reasoning, agent planning, knowledge work and design.
  • Default model for the Free and Pro plans of claude.ai and for Claude Cowork.
  • Better resistance to prompt injections; adaptive and extended thinking, context compaction.
  • According to the vendor, Claude Code users prefer it to Sonnet 4.5 about 70% of the time, and to Opus 4.5 in 59% of cases.

### Specifications - Context window: 1M tokens in beta at launch (out of beta on March 13, 2026). - Maximum output: not specified in the post. Knowledge cutoff: not disclosed. - Availability: all Claude plans, Claude Code, API and major clouds. API identifier: `claude-sonnet-4-6`. ### Pricing | Item | Pricing (USD per million tokens) | | --- | --- | | Input | 3 | | Output | 15 | | Cache write 5 min / 1 h | 3.75 / 6 | | Cache read | 0.30 | | Batch (input / output) | 1.50 / 7.50 | ### Benchmarks Scores published by the vendor (the post mostly refers to charts). | Evaluation | Score | | --- | --- | | SWE-bench Verified (with prompt modification) | 80.2% | | ARC-AGI-2 (high effort) | 60.4% | | Insurance benchmark (Pace) | 94% | ### Key takeaways Sonnet 4.6 brings the mid-range tier closer to Opus level on coding and computer use, with no price increase.

Links

Tags : LLM, Anthropic, Claude, Sonnet, Agents, Usage d'ordinateur