HELM VHELM - Mmmu - Art Theory: leaderboard
Metric: Prefix Quasi-Exact Match (%). Source: crfm.stanford.edu. 48 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 (2025-04-14) | 93.33 |
| 2 | Gemini 2.5 Pro (Preview 03-25) | 93.33 |
| 3 | GPT-4o (2024-05-13) | 93.33 |
| 4 | GPT-4 Turbo | 93.33 |
| 5 | O4 Mini (2025-04-16) | 93.33 |
| 6 | O3 (2025-04-16) | 93.33 |
| 7 | GPT-4.5 | 93.33 |
| 8 | GPT-4o (2024-11-20) | 93.33 |
| 9 | Gemini 2.0 Pro (Preview 02-05) | 93.33 |
| 10 | Gemini 2.0 Flash Lite | 93.33 |
| 11 | GPT-4.1 Mini | 90 |
| 12 | Gemini 2.0 Flash | 90 |
| 13 | GPT-4o (2024-08-06) | 90 |
| 14 | Gemini 2.0 Flash (Preview) | 90 |
| 15 | Claude 3.5 Sonnet (20241022) | 86.67 |
Interactive version: theaggregate.ai/benchmark?slug=helm-vhelm-mmmu-art-theory · How It Works · Data refreshed daily, snapshot 2026-09-19.