HELM VHELM - Crossmodal 3600: leaderboard
The "Crossmodal 3600" 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: Prometheus Vision rating. Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 48 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O4 Mini (2025-04-16) | 3.96 | #181 |
| 2 | GPT-4 Turbo | 3.72 | #406 |
| 3 | GPT-4.1 Mini | 3.7 | #351 |
| 4 | Claude 3.5 Sonnet (20241022) | 3.65 | #291 |
| 5 | Gemini 2.5 Pro (Preview 03-25) | 3.65 | #124 |
| 6 | O3 (2025-04-16) | 3.61 | #123 |
| 7 | GPT-4.1 (2025-04-14) | 3.53 | #250 |
| 8 | Claude 3.5 Sonnet (20240620) | 3.52 | #298 |
| 9 | GPT-4o Mini (2024-07-18) | 3.52 | #561 |
| 10 | O1 (2024-12-17) | 3.49 | #152 |
| 11 | GPT-4.5 | 3.38 | #276 |
| 12 | Gemini 1.5 Pro (002) | 3.36 | #350 |
| 13 | Gemini 2.0 Pro (Preview 02-05) | 3.35 | #305 |
| 14 | GPT-4.1 Nano | 3.31 | #735 |
| 15 | GPT-4o (2024-11-20) | 3.3 | #378 |
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-crossmodal-3600 · How It Works · Data refreshed daily, snapshot 2026-10-23.