Models
Claude Sonnet 4.6
| Provider | Anthropic |
|---|---|
| Context Window | 200K tokens |
| Parameters | Unknown (estimated ~70B) |
| Input Pricing | $3/1M |
| Output Pricing | $15/1M |
| API Available | ✓ Yes |
| Open Source | ✗ No |
Claude Sonnet 4.6 is Anthropic’s flagship mid-tier model, designed to deliver near-Opus quality at Haiku-approaching speeds. It is the recommended model for the vast majority of production use cases.
Capabilities
- Extended thinking: Supports long-horizon reasoning with a scratchpad mechanism for complex problems
- Tool use: Reliable function calling, code execution, and computer use
- Vision: Processes images, charts, diagrams, and screenshots
- Code: Best-in-class at refactoring, debugging, and explaining complex codebases
- Writing: Long-form content, summaries, translation across 50+ languages
Performance Summary
On MMLU: 91.2% | On HumanEval: 88.4% | On MATH: 87.1%
Best For
Production APIs, agentic workflows, coding assistants, document analysis, customer support automation, and any application where quality matters but Opus-level cost is prohibitive.
Context Window Notes
The 200K context window (~150,000 words) handles entire books, large codebases, or lengthy document collections in a single prompt. Recall quality remains high throughout the window.