OpenVLM MME — leaderboard

OpenCompass OpenVLM evaluation of MME: perception and cognition tests for multimodal models covering OCR, object existence, position, color, celebrity, artwork, commonsense, numerical calculation, and text translation.

Metric: Overall Score. Source: huggingface.co. Status: saturated. 235 models tracked.

Top models

#ModelScore
1InternVL3-78B2538.6
2Qwen 2 VL 72B2504.2
3InternVL3-38B2500.7
4InternVL2.5-78B2467.6
5InternVL3-14B2462.3
6InternVL3-8B2422
7Gemini 2.0 Flash2386.1
8GPT-4.1 (2025-04-14)2337.6
9Ovis2-8B2326.6
10Qwen 2 VL 7B2276.3
11MiniCPM-V-2.62268.7
12InternVL2-8B2215.1
13Grok 2 (1212)2154.7
14Gemini 1.0 Pro2148.9
15Gemma 3 27B2148.5

Interactive version: theaggregate.ai/benchmark?slug=openvlm-mme · How the rankings work · Data refreshed daily, snapshot 2026-07-22.