HELM VHELM - Blink Perception: leaderboard
The "Blink Perception" 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) | 84.61 | #123 |
| 2 | Llama 4 Maverick Instruct FP8 | 84.43 | #412 |
| 3 | GPT-4.5 | 84.28 | #276 |
| 4 | Gemini 2.5 Pro (Preview 03-25) | 83.86 | #124 |
| 5 | O4 Mini (2025-04-16) | 83.27 | #181 |
| 6 | O1 (2024-12-17) | 82.76 | #152 |
| 7 | GPT-4.1 (2025-04-14) | 82.48 | #250 |
| 8 | Gemini 2.0 Flash (Preview) | 81.83 | #366 |
| 9 | GPT-4o (2024-05-13) | 81.25 | #326 |
| 10 | Gemini 2.0 Flash Lite | 80.62 | #453 |
| 11 | Gemini 1.5 Pro (002) | 80.35 | #350 |
| 12 | Gemini 2.0 Flash | 80.24 | #344 |
| 13 | Claude 3 Sonnet (20240229) | 79.85 | #681 |
| 14 | Claude 3 Opus (20240229) | 79.4 | #413 |
| 15 | Gemini 2.0 Pro (Preview 02-05) | 79 | #305 |
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-blink-perception · How It Works · Data refreshed daily, snapshot 2026-10-23.