Key capabilities
- OpenAI-compatible — Works as a drop-in replacement with the OpenAI SDK
- Extended thinking — Step-by-step internal reasoning enabled by default
- Balanced performance — Strong reasoning at Sonnet-level speed and cost
- Advanced coding — Excels at code generation, debugging, and refactoring
- Long context — Handles large documents and multi-turn conversations
- Streaming — Supports real-time token streaming via SSE
Quick example
Thinking sampling compatibility: For reliable requests, omit
temperature and top_k. Anthropic accepts the unchanged default temperature: 1 and top_p values from 0.95 to 1. A non-default temperature, any top_k, or a top_p below 0.95 returns a 400 error while Thinking is enabled. See Anthropic’s Thinking limits and API usage primer.Parameters
API Reference
View the interactive API playground for Claude Sonnet 4 Thinking.