Benchmarks / Agentic
CUA-bench
Computer-use agents playing commercial video games through keyboard and mouse input, with half the titles held out.
CUA-bench
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | GPT-6 Astra | max | 19.2% |
| 2 | Opus 5.5 | max | 14.0% |
| 3 | Fable 5.1 | max | 13.2% |
| 4 | Opus 5 | max | 9.0% |
| 5 | GPT-5.6 Sol | max | 8.3% |
| 6 | Wonder | max | 7.2% |
| 7 | Muse Spark 1.3 | max | 5.8% |
| 8 | Gemini 4 | high | 4.8% |
| 9 | Gemini 3.8 Flash | high | 4.2% |