HELM VHELM - Gqa: leaderboard
The "Gqa" 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 | O3 (2025-04-16) | 64.6 | #124 |
| 2 | Gemini 2.5 Pro (Preview 03-25) | 64.1 | #125 |
| 3 | GPT-4.5 | 63.7 | #276 |
| 4 | Gemini 2.0 Flash (Preview) | 63.2 | #369 |
| 5 | Gemini 2.0 Flash Lite | 63 | #455 |
| 6 | O4 Mini (2025-04-16) | 62.5 | #182 |
| 7 | Gemini 2.0 Pro (Preview 02-05) | 62.1 | #306 |
| 8 | GPT-4o (2024-05-13) | 61.9 | #328 |
| 9 | Gemini 2.0 Flash | 61.8 | #345 |
| 10 | GPT-4.1 (2025-04-14) | 61.4 | #252 |
| 11 | O1 (2024-12-17) | 61.2 | #153 |
| 12 | GPT-4o (2024-11-20) | 59.2 | #379 |
| 13 | GPT-4o (2024-08-06) | 59.1 | #337 |
| 14 | GPT-4.1 Mini | 58.3 | #358 |
| 15 | Claude 3.5 Sonnet (20240620) | 55.1 | #300 |
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-gqa · How It Works · Data refreshed daily, snapshot 2026-10-24.