Claude Sonnet 5.5 is the everyday complement to Opus 5.5: strongest at well-scoped tasks such as fixing bugs and producing polished documents, slides and spreadsheets. It is a large step up from Sonnet 5 on agentic work (Terminal-Bench 4.0 70.6% vs 10.3%; OSWorld 2.1 80.1% vs 57.0%) and matches Opus 5.5 on knowledge-work evaluations, while staying at $2/$10 per million tokens. Adaptive thinking is on by default and can be limited to between tool calls. It accepts text and images with a 1M-token context window and up to 128K output tokens.
Reliable instruction following
Strong tool use and function calling
Excellent code generation and debugging
1M token context window
Strong reasoning with extended thinking
Accepts image inputs
Broad multilingual support
Strong creative writing ability
| Benchmark | Score | Type | Recorded |
|---|---|---|---|
| Humanity's Last Exam | 55.0 | accuracy | 6d ago |
| TerminalBench | 63.6 | accuracy | 6d ago |
| LCR | 82.7 | accuracy | 6d ago |
| SciCode | 61.0 | accuracy | 6d ago |