Benchmark
OSWorld
multimodal
general
vision
multimodal
OSWorld: The first-of-its-kind scalable, real computer environment for multimodal agents, supporting task setup, execution-based evaluation, and interactive learning across Ubuntu, Windows, and macOS with 369 computer tasks involving real web and desktop applications, OS file I/O, and multi-application workflows
语言EN
满分1
参评模型7
模型排名
| 名次 | 模型 | 机构 | 分数 | 来源 |
|---|---|---|---|---|
| 1 | GPT-5.6 Sol | OpenAI | 62.6 | 来源 ↗ |
| 2 | Claude Sonnet 4.5 | Anthropic | 61.4 | 来源 ↗ |
| 3 | Claude Haiku 4.5 | Anthropic | 50.7 | 来源 ↗ |
| 4 | GPT-5.6 Terra | OpenAI | 50.2 | 来源 ↗ |
| 5 | GPT-5.6 Luna | OpenAI | 45.6 | 来源 ↗ |
| 6 | Qwen2.5 VL 72B Instruct | Alibaba Cloud / Qwen Team | 8.8 | 来源 ↗ |
| 7 | Qwen2.5 VL 32B Instruct | Alibaba Cloud / Qwen Team | 5.9 | 来源 ↗ |