HELM VHELM - Mmmu - Geography: leaderboard
VHELM (Stanford CRFM HELM for vision-language models, run v2.1.2) MMMU Geography: college-level multiple-choice questions over images from the MMMU validation set for this subject; 27 questions, prefix quasi-exact match. Percent correct, 0-100, higher is better; a small sample.
Metric: Prefix Quasi-Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O3 (2025-04-16) | 81.48 | #125 |
| 2 | Gemini 2.0 Flash Lite | 74.07 | #455 |
| 3 | Claude 3.5 Sonnet (20241022) | 70.37 | #302 |
| 4 | GPT-4o (2024-08-06) | 70.37 | #343 |
| 5 | O1 (2024-12-17) | 70.37 | #157 |
| 6 | GPT-4.5 | 70.37 | #278 |
| 7 | Gemini 2.0 Flash | 66.67 | #351 |
| 8 | Claude 3.5 Sonnet (20240620) | 66.67 | #308 |
| 9 | Claude 3 Opus (20240229) | 66.67 | #428 |
| 10 | Gemini 2.0 Pro (Preview 02-05) | 66.67 | #318 |
| 11 | Gemini 2.5 Pro (Preview 03-25) | 62.96 | #126 |
| 12 | GPT-4.1 (2025-04-14) | 62.96 | #257 |
| 13 | Claude 3 Haiku (20240307) | 62.96 | #780 |
| 14 | Gemini 1.5 Pro (002) | 59.26 | #368 |
| 15 | GPT-4o Mini (2024-07-18) | 59.26 | #568 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-vhelm-mmmu-geography · How It Works · Data refreshed daily, snapshot 2026-11-05.