HELM VHELM - Blink Perception - Object Localization: leaderboard
VHELM (Stanford CRFM HELM for vision-language models, run v2.1.2) BLINK Object Localization task: which bounding box correctly localizes an object, one of BLINK's perception tasks that humans solve at a glance, asked as multiple choice; 122 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 | Claude 3 Opus (20240229) | 100 | #428 |
| 2 | GPT-4o (2024-11-20) | 95.9 | #392 |
| 3 | GPT-4.1 (2025-04-14) | 91.8 | #257 |
| 4 | GPT-4o (2024-05-13) | 90.98 | #337 |
| 5 | Claude 3 Sonnet (20240229) | 90.98 | #694 |
| 6 | Claude 3 Haiku (20240307) | 88.52 | #780 |
| 7 | GPT-4.5 | 88.52 | #278 |
| 8 | Gemini 1.5 Pro (002) | 82.79 | #368 |
| 9 | Gemini 1.5 Flash (002) | 82.79 | #587 |
| 10 | GPT-4o (2024-08-06) | 82.79 | #343 |
| 11 | O3 (2025-04-16) | 82.79 | #125 |
| 12 | Gemini 2.0 Flash (Preview) | 82.79 | #376 |
| 13 | Claude 3.5 Sonnet (20240620) | 81.15 | #308 |
| 14 | O1 (2024-12-17) | 81.15 | #157 |
| 15 | Gemini 2.0 Flash | 80.33 | #351 |
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-object-localization · How It Works · Data refreshed daily, snapshot 2026-11-05.