OMLAB Open Agent Multimodal Leaderboard

Open Agent Leaderboard multimodal track comparing visual-agent configurations by score, pass rate, and token usage.

9rows
scoreprimary metric
2026-05-06sampled

Metadata

ID
omlab_open_agent_multimodal
Category
Multimodal
Release
2025-01-09

Metrics

Score, Pass Rate, Total Input Tokens (lower is better), Total Output Tokens (lower is better), All Tokens (lower is better)

Latest Results

Rows are parsed from OMLAB's public Open Agent Leaderboard multimodal CSV. Source agent and VLM strings are preserved as the evaluated configuration.

Rank Subject Score Model Match Provenance Sampled
1 ZoomEye + Qwen2.5-VL-72B-Instruct 51.56 — Imported 2026-05-06
2 ZoomEye + Qwen2.5-VL-7B-Instruct 48.06 — Imported 2026-05-06
3 IO + Qwen2.5-VL-72B-Instruct 44.47 — Imported 2026-05-06
4 ZoomEye + InternVL2.5-8B 43.42 — Imported 2026-05-06
5 IO + InternVL2.5-8B 42.95 — Imported 2026-05-06
6 IO + Qwen2.5-VL-7B-Instruct 42.86 — Imported 2026-05-06
7 ZoomEye + Llava-v1.5-7B 31.60 — Imported 2026-05-06
8 IO + Llava-v1.5-7B 24.79 — Imported 2026-05-06
9 V* + seal_vqa & seal_vsm 15.14 — Imported 2026-05-06