HELM VHELM - Blink Perception - Visual Similarity: leaderboard
VHELM (Stanford CRFM HELM for vision-language models, run v2.1.2) BLINK Visual Similarity task: which image is most similar to the reference, one of BLINK's perception tasks that humans solve at a glance, asked as multiple choice; 135 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 | Llama 4 Maverick Instruct FP8 | 97.78 | #424 |
| 2 | Gemini 2.0 Flash (Preview) | 97.78 | #376 |
| 3 | Gemini 2.0 Flash | 94.81 | #351 |
| 4 | GPT-4.1 (2025-04-14) | 94.81 | #257 |
| 5 | Gemini 2.0 Flash Lite | 94.07 | #455 |
| 6 | Gemini 2.5 Pro (Preview 03-25) | 93.33 | #126 |
| 7 | GPT-4o (2024-05-13) | 93.33 | #337 |
| 8 | O1 (2024-12-17) | 92.59 | #157 |
| 9 | GPT-4.5 | 92.59 | #278 |
| 10 | Gemini 1.5 Pro (002) | 91.85 | #368 |
| 11 | Claude 3 Opus (20240229) | 91.85 | #428 |
| 12 | O4 Mini (2025-04-16) | 90.37 | #183 |
| 13 | O3 (2025-04-16) | 90.37 | #125 |
| 14 | Gemini 1.5 Flash (002) | 89.63 | #587 |
| 15 | GPT-4o (2024-11-20) | 88.89 | #392 |
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-blink-perception-visual-similarity · How It Works · Data refreshed daily, snapshot 2026-11-05.