HELM VHELM - Exams V: leaderboard
The "Exams V" subtask of HELM VHELM. Stanford HELM's VHELM project, the holistic evaluation of vision-language models: image-grounded question answering, OCR, chart and diagram reading, spatial reasoning, robustness, bias, fairness and toxicity scenarios scored under one ha...
Metric: PEM. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 03-25) | 80.68 | #124 |
| 2 | O3 (2025-04-16) | 77.58 | #123 |
| 3 | O4 Mini (2025-04-16) | 75.83 | #181 |
| 4 | O1 (2024-12-17) | 72.38 | #152 |
| 5 | GPT-4.5 | 66.69 | #276 |
| 6 | Gemini 2.0 Flash | 66.05 | #344 |
| 7 | Gemini 2.0 Flash (Preview) | 65.83 | #366 |
| 8 | GPT-4.1 (2025-04-14) | 64.18 | #250 |
| 9 | GPT-4.1 Mini | 61.56 | #351 |
| 10 | GPT-4o (2024-08-06) | 61.36 | #335 |
| 11 | GPT-4o (2024-05-13) | 60.82 | #326 |
| 12 | GPT-4o (2024-11-20) | 60.54 | #378 |
| 13 | Gemini 2.0 Flash Lite | 59.5 | #453 |
| 14 | Gemini 2.0 Pro (Preview 02-05) | 58.9 | #305 |
| 15 | Claude 3.5 Sonnet (20240620) | 57.68 | #298 |
No result here: #4 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-vhelm-exams-v · How It Works · Data refreshed daily, snapshot 2026-10-23.