Diagram-MMU - Diagram QA: leaderboard
Metric: Accuracy (%). Source: vi-ocean.github.io. 11 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 86.29 |
| 2 | Qwen 3.5 397B A17B | 83.42 |
| 3 | GPT-5.2 | 81.67 |
| 4 | Kimi K2.5 | 79.74 |
| 5 | Seed 2.0 Pro | 75.9 |
| 6 | Qwen 3 VL 235B A22B | 63.6 |
| 7 | InternVL3-38B | 56.39 |
Interactive version: theaggregate.ai/benchmark?slug=diagram-mmu-diagram-qa · How It Works · Data refreshed daily, snapshot 2026-09-05.