LLM Stats (OSWorld-Verified): leaderboard
Human-verified subset of OSWorld with confirmed task completability, testing autonomous computer use in real OS environments.
Metric: Score (%). Source: llm-stats.com. Saturation forecast: Around January 2027. 28 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.8 Max | 86.1 | #35 |
| 2 | Claude Fable 5 | 85 | #16 |
| 3 | Qwen 3.8 27B | 84.3 | #105 |
| 4 | Claude Opus 4.8 | 83.4 | #37 |
| 5 | Gemini 3.6 Flash | 83 | #51 |
| 6 | MiMo-V2.6-Pro | 82 | #42 |
| 7 | MiMo-V2.6-Flash | 80.8 | #70 |
| 8 | Muse Spark 1.1 | 80.8 | #52 |
| 9 | Muse Spark Open | 80.4 | #54 |
| 10 | Claude Mythos Preview | 79.6 | #4 |
| 11 | Gemini 3.5 Flash | 78.4 | #60 |
| 12 | Claude Opus 4.7 | 78 | #49 |
| 13 | GPT-5.4 | 75 | #79 |
| 14 | Gemini 3.5 Flash Lite | 74 | #156 |
| 15 | Qwen 3.7 Plus | 73.3 | #92 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=llm-stats-osworld-verified · How It Works · Data refreshed daily, snapshot 2026-10-20.