HELM VHELM - Crossmodal 3600 - Maori Locale: leaderboard
VHELM (Stanford CRFM HELM for vision-language models, run v2.1.2) Crossmodal-3600 captioning of the 100 images from the Maori locale: the model writes a short English caption, rated 1-5 by a Prometheus-Vision judge against human reference captions. Mean rating on the 1-5 scale, higher is better.
Metric: Prometheus Vision Rating (1-5). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O4 Mini (2025-04-16) | 4.09 | #183 |
| 2 | Claude 3.5 Sonnet (20241022) | 3.93 | #302 |
| 3 | O3 (2025-04-16) | 3.79 | #125 |
| 4 | Gemini 2.5 Pro (Preview 03-25) | 3.78 | #126 |
| 5 | GPT-4.1 (2025-04-14) | 3.74 | #257 |
| 6 | GPT-4.1 Mini | 3.73 | #363 |
| 7 | GPT-4 Turbo | 3.69 | #417 |
| 8 | O1 (2024-12-17) | 3.67 | #157 |
| 9 | GPT-4o Mini (2024-07-18) | 3.62 | #568 |
| 10 | Gemini 2.0 Pro (Preview 02-05) | 3.57 | #318 |
| 11 | Claude 3 Opus (20240229) | 3.52 | #428 |
| 12 | Claude 3 Haiku (20240307) | 3.46 | #780 |
| 13 | GPT-4.5 | 3.45 | #278 |
| 14 | Gemini 1.5 Pro (002) | 3.4 | #368 |
| 15 | GPT-4o (2024-11-20) | 3.39 | #392 |
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-crossmodal-3600-maori-locale · How It Works · Data refreshed daily, snapshot 2026-11-05.