HELM VHELM - Exams V - Chinese (Natural Science): leaderboard
VHELM (Stanford CRFM HELM for vision-language models, run v2.1.2) Exams-V questions in Chinese on natural science: school exam multiple-choice questions rendered as images that combine text, formulas and figures; 285 questions, prefix quasi-exact match. Percent correct, 0-100, higher is better.
Metric: Prefix Quasi-Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O4 Mini (2025-04-16) | 74.74 | #183 |
| 2 | O3 (2025-04-16) | 74.74 | #125 |
| 3 | O1 (2024-12-17) | 68.07 | #157 |
| 4 | Gemini 2.0 Flash (Preview) | 64.91 | #376 |
| 5 | Gemini 2.5 Pro (Preview 03-25) | 62.46 | #126 |
| 6 | Gemini 2.0 Flash | 61.75 | #351 |
| 7 | Gemini 2.0 Flash Lite | 58.95 | #455 |
| 8 | Gemini 2.0 Pro (Preview 02-05) | 55.79 | #318 |
| 9 | GPT-4 Turbo | 50.88 | #417 |
| 10 | GPT-4.1 (2025-04-14) | 48.77 | #257 |
| 11 | Gemini 1.5 Flash (002) | 47.37 | #587 |
| 12 | Claude 3.5 Sonnet (20240620) | 47.37 | #308 |
| 13 | Gemini 1.5 Pro (002) | 46.32 | #368 |
| 14 | Claude 3 Opus (20240229) | 45.61 | #428 |
| 15 | GPT-4.5 | 45.61 | #278 |
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-exams-v-chinese-natural-science · How It Works · Data refreshed daily, snapshot 2026-11-05.