cua-speedrun - OSWorld (50 Tasks): leaderboard
Metric: Task success (%), one run per task in the cua-speedrun Modal harness. Source: cuaspeedrun.com. Saturation forecast: Around January 2027. 53 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Opus 5 (High) | 91.62 |
| 2 | GPT-6 Astra (xHigh) | 91.62 |
| 3 | Gemini 3.8 Flash (Medium) | 91.62 |
| 4 | Gemini 3.8 Flash (Low) | 91.62 |
| 5 | Kimi K3 (Max) | 89.62 |
| 6 | Gemini 3.7 Flash (High) | 89.62 |
| 7 | Claude Sonnet 5 (High) | 89.62 |
| 8 | GPT-6 Astra (High) | 89.62 |
| 9 | GPT-6 Astra (Medium) | 89.62 |
| 10 | Gemini 3.7 Flash (Medium) | 89.62 |
| 11 | Gemini 3.8 Flash (High) | 87.62 |
| 12 | GPT-6 Astra (Low) | 87.62 |
| 13 | Claude Opus 5 (Low) | 87.62 |
| 14 | Gemini 3.7 Flash (Low) | 87.62 |
| 15 | GPT-5.6 Sol (xHigh) | 81.62 |
Interactive version: theaggregate.ai/benchmark?slug=cua-speedrun-osworld-50-tasks · How It Works · Data refreshed daily, snapshot 2026-10-03.