Claude Sonnet 4.6 by Anthropic: default model on claude.ai, near-Opus intelligence at Sonnet pricing, major computer use upgrade, adaptive thinking, and 1M token context window in beta.
Capabilities, design details, and architectural traits
Claude Sonnet 4.6 is the default model across Claude.ai and Claude Cowork for Free and Pro users. Its defining identity is closing the gap to Opus-class intelligence while remaining at Sonnet pricing - to the point where early users in Claude Code preferred it over Opus 4.5, the previous generation flagship, 59% of the time.
| Trait | What it means for Sonnet 4.6 specifically |
|---|---|
| Default model on claude.ai and Cowork | Replaces prior Sonnet as the default for all Free and Pro users across claude.ai and Claude Cowork at launch |
| Computer use - major generational step | Officially described as a "major improvement" over Sonnet 4.5; prompt injection resistance matches Opus 4.6, upgraded from experimental toward production use |
| Both adaptive and extended thinking supported | Sonnet 4.6 supports both thinking: {type: "adaptive"} and the older budget_tokens extended thinking mode (deprecated but still functional), unlike Opus 4.7+ which dropped budget_tokens entirely |
| 1M token context window in beta | Available in beta on the API; Anthropic documents that the model reasons effectively across this full context, not just ingests it |
| Strong performance with thinking off | Officially documented to deliver strong performance even with extended thinking disabled - explicitly recommended to explore across the effort spectrum |
| Context compaction in beta | Automatically summarizes older context as conversations approach limits, extending effective length for long agentic sessions |
Anthropicexplicitly highlights prompt injection resistance as a safety advance specific to Sonnet 4.6 in its computer use context. The system card confirms Sonnet 4.6 is a major improvement over Sonnet 4.5 on this vector, and now performs similarly to Opus 4.6 - a documented safety parity that was not present in prior Sonnet models.
Independent evaluations · Artificial Analysis
Evaluate specifications, pricing, and independent benchmark indices
| Model Details | |||
|---|---|---|---|
| General Info | |||
| Provider | Anthropic | Anthropic | Anthropic |
| Release Date | February 17, 2026 | July 24, 2026 | June 9, 2026 |
| Knowledge Cutoff | - | May 2026 | - |
| Context & Limits | |||
| Context Window | 1M | 1M | 1M |
| Pricing (per 1M tokens) | |||
| Input Pricing | $3 Best Input Pricing | $5 | $10 |
| Output Pricing | $15 Best Output Pricing | $25 | $50 |
| Modalities | |||
| Inputs | textimagefile | textimage | textimagefile |
| Outputs | text | text | text |
| Benchmarks (0-100) | |||
| Intelligence Index | 48.4 | 63.1 Best Intelligence Index | 62.1 |
| Coding Index | 63.0 | 78.0 Best Coding Index | 76.5 |
| Agentic Index | 42.1 | 59.2 Best Agentic Index | 56.6 |
Graduate-level reasoning and expert Q&A evaluation.
Extremely difficult logical reasoning and knowledge.
Logical reasoning over long context windows.
Independent evaluation data provided by Artificial Analysis. To view the latest benchmarks and full details, visit their official site.
