We have added Claude Opus 5.5 to OpenCheese.
Anthropic describes Opus 5.5 as its most capable Opus model for coding, agents, and knowledge work. It is designed for large codebases, complex document creation, financial analysis, computer use, and long running workflows that require careful planning and minimal supervision.
The model also focuses on efficiency. Anthropic estimates that typical workloads billed by tokens cost around 40% less to run than with Opus 5. Opus 5.5 uses fewer tokens per task, communicates more clearly, and follows writing and formatting rules more consistently. A Fast mode with up to 2.5 times greater speed is also available through Anthropic.
Benchmark comparison
| Benchmark | Opus 5.5 | Opus 5 | Fable 5.1 | GPT 6 Astra | Grok 4.7 |
|---|---|---|---|---|---|
| Terminal Bench 4.0 | 65.4 | 52.3 | 49.5 | 55.7 | 34.9 |
| AutomationBench | 59.3 | 52.0 | 52.1 | 62.1 | 53.1 |
| OpenClaw Striker | 69.2 | 66.4 | 66.6 | 70.4 | 58.0 |
| GDPval AA 2.1 | 1697 | 1629 | 1589 | 1592 | 1673 |
The results show a particularly large improvement in software engineering. Opus 5.5 scores 65.4 on Terminal Bench 4.0, compared with 52.3 for Opus 5. It also leads the selected models on GDPval AA 2.1. These figures were published by Anthropic, and actual performance may vary by workflow and configuration.
Read more on the official Anthropic page.
You can try Claude Opus 5.5 at opencheese.dev right now!