Diagram-MMU - Diagram QA: leaderboard
The question-answering track of Diagram-MMU (arXiv 2608.12262): accuracy % on questions about scientific diagrams that can only be answered by reading the figure itself rather than its caption.
Metric: Accuracy (%). Source: vi-ocean.github.io. Saturation forecast: Around January 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 86.29 | #58 |
| 2 | Qwen 3.5 397B A17B | 83.42 | #146 |
| 3 | GPT-5.2 | 81.67 | #110 |
| 4 | Kimi K2.5 | 79.74 | #144 |
| 5 | Seed 2.0 Pro | 75.9 | #118 |
| 6 | Claude Opus 4.6 | 68.75 | #63 |
| 7 | Qwen 3 VL 235B A22B | 63.6 | #238 |
| 8 | InternVL3-38B | 56.39 | #405 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=diagram-mmu-diagram-qa · How It Works · Data refreshed daily, snapshot 2026-10-20.