Claude Sonnet 5.5
modelYour notes
The second model in the Claude 5.5 family, six days after Opus 5.5. Anthropic calls it a clear upgrade over Sonnet 5 that runs more than 30% faster and costs up to 30% less for most work at an unchanged price of $2 / $10 per million input/output tokens, because it needs far fewer tokens per task. Artificial Analysis scores it 55 on Intelligence Index v5.0.1 at max effort (54 on v5.0). That puts it behind Opus 5.5 (59) and GPT-6 Astra (57), level with GPT-6.1 Sol (55), and ahead of Fable 5.1 (49); on v4.3.2 it read 56, #2 overall behind Opus 5.5. Anthropic positions it for well-scoped everyday tasks, bug fixing and polished documents, and says Opus 5.5 remains clearly stronger at open-ended work that needs sustained judgment. API ID claude-sonnet-5-5; 1M-token context, 128K output.
The jump over Sonnet 5 is largest on agentic work: Terminal-Bench 4.0 70.6% against 10.3% (and above Opus 5.5's 66.4%), GDPval-AA v2.1 Elo 1844 against 1449, AA-Briefcase v1.1 Elo 1811 against 1359, OSWorld 2.1 80.1% against 57.0%, CursorBench 4.0 55.5% against 34.1%, and Humanity's Last Exam with tools 64.5% against 54.9%. It is the first Sonnet to finish Pokémon Red from screenshots alone and the first to ship with the cyber safeguards and model fallbacks Anthropic built for its most capable models, since its cyber capability is comparable to Opus 5's. Haiku 5.5 followed on October 7.
Model Details
Benchmark Scores
| Benchmark | Score | Mode |
|---|---|---|
| Terminal-Bench 4.0 | 70.6% | — |
| GDPval-AA v2.1 (Elo) | 1844 | — |
| OSWorld 2.1 | 80.1% | partial credit |
| CursorBench 4.0 | 55.5% | — |
| Humanity's Last Exam | 64.5% | with tools |
Variants
| Name | Parameters | Notes |
|---|---|---|
| Claude Sonnet 5.5 (xhigh) | — | AAII v5.0 50 (v4.3.2 52) |
| Claude Sonnet 5.5 (high) | — | AAII v5.0 41 (v4.3.2 47) |