OMLAB Open Agent Multimodal Leaderboard — leaderboard

Open Agent Leaderboard multimodal track where visual agents interpret screen state, choose actions, and complete interactive tasks.

Metric: Score. Source: huggingface.co. 9 models tracked.

Top models

#ModelScore
1Qwen2.5-VL-72B-Instruct (ZoomEye)51.56
2Qwen2.5-VL-7B-Instruct (ZoomEye)48.06
3Qwen2.5-VL-72B-Instruct (IO)44.47
4InternVL2.5-8B (ZoomEye)43.42
5InternVL2.5-8B (IO)42.95
6Qwen2.5-VL-7B-Instruct (IO)42.86
7Llava-v1.5-7B (ZoomEye)31.6
8Llava-v1.5-7B (IO)24.79
9seal_vqa & seal_vsm (V*)15.14

Interactive version: theaggregate.ai/benchmark?slug=omlab-open-agent-multimodal-leaderboard · How the rankings work · Data refreshed daily, snapshot 2026-07-22.