FrontierCUA avg@1
Agents
Partial-credit companion to FrontierCUA pass@1 on the same 15 enterprise computer-use workflows: the unweighted mean over workflows of each workflow's mean graded attempt score, the weighted share of rubric criteria met on a 0-1 scale. The leaderboard uses it to break pass@1 ties.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Opus 5.5 | 0.7 |
| Claude Fable 5.1 | 0.7 |
| GPT-6.1 Sol | 0.6 |
| Claude Fable 5 | 0.6 |
| GPT-6 Astra | 0.6 |
| GPT-6 Sol | 0.5 |
| Sonnet 5 | 0.4 |
| GPT-5.6 Sol | 0.4 |
| Kimi K3 | 0.4 |
| DeepSeek V4.1 Flash | 0.3 |
| GLM-5.3 Flash | 0.3 |
| Grok 4.7 | 0.3 |
| MiMo-V2.6-Pro | 0.2 |