multimodal benchmark
OSWorld-Verified is a verified subset of OSWorld, a scalable real computer environment for multimodal agents supporting task setup, execution-based evaluation, and interactive learning across Ubuntu, Windows, and macOS.
Updated Aug 11, 2026
Higher score ranks better on this benchmark.
| 01 | AC | 86.1% | 100.0% | 23 | C | |
| 02 | AN | 85.0% | 95.5% | 23 | C | |
| 03 | AN | 83.4% | 90.9% | 23 | C | |
| 04 | GO | 83.0% | 86.4% | 23 | C | |
| 05 | AN | 81.2% | 81.8% | 23 | C | |
| 06 | ME | 80.8% | 77.3% | 23 | C | |
| 07 | AN | 79.6% | 72.7% | 23 | C | |
| 08 | OP | 78.7% | 68.2% | 23 | C | |
| 09 | GO | 78.4% | 63.6% | 23 | C | |
| 10 | AN | 78.0% | 59.1% | 23 | C | |
| 11 | OP | 75.0% | 54.5% | 23 | C | |
| 12 | GO | 74.0% | 50.0% | 23 | C | |
| 13 | AC | 73.3% | 45.5% | 23 | C | |
| 14 | MA | 73.1% | 40.9% | 23 | C | |
| 15 | OP | 72.1% | 36.4% | 23 | C | |
| 16 | MI | 70.1% | 31.8% | 23 | C | |
| 17 | ME | 65.9% | 27.3% | 23 | C | |
| 18 | OP | 64.7% | 22.7% | 23 | C | |
| 19 | AC | 62.5% | 18.2% | 23 | C | |
| 20 | AC | 58.0% | 13.6% | 23 | C | |
| 21 | AC | 56.2% | 9.1% | 23 | C | |
| 22 | AC | 54.5% | 4.5% | 23 | C | |
| 23 | OP | 39.0% | 0.0% | 23 | C |
A closer view of the leading scores on this benchmark.
The leading models and scores on this benchmark.
What OSWorld-Verified measures and how its scores work.
OSWorld-Verified is a verified subset of OSWorld, a scalable real computer environment for multimodal agents supporting task setup, execution-based evaluation, and interactive learning across Ubuntu, Windows, and macOS.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
Common questions about OSWorld-Verified.
Qwen3.8 Max is currently ranked first with 86.1%.
OSWorld-Verified is a verified subset of OSWorld, a scalable real computer environment for multimodal agents supporting task setup, execution-based evaluation, and interactive learning across Ubuntu, Windows, and macOS.
Yes. Higher values rank better for this benchmark.
23 model results are currently shown.
Yes. This benchmark can contribute to the current LLMBoard capability score.