HELM VHELM - Blink Reasoning - Jigsaw: leaderboard
VHELM (Stanford CRFM HELM for vision-language models, run v2.1.2) BLINK Jigsaw task: which patch completes a jigsaw image, one of BLINK's perception tasks that humans solve at a glance, asked as multiple choice; 150 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 Sonnet (20240229) | 100 | #694 |
| 2 | Claude 3 Opus (20240229) | 100 | #428 |
| 3 | Claude 3 Haiku (20240307) | 96.67 | #780 |
| 4 | GPT-4o Mini (2024-07-18) | 96 | #568 |
| 5 | Llama 4 Maverick Instruct FP8 | 95.33 | #424 |
| 6 | Gemini 2.0 Flash (Preview) | 95.33 | #376 |
| 7 | Claude 3.5 Sonnet (20240620) | 94.67 | #308 |
| 8 | GPT-4o (2024-05-13) | 94.67 | #337 |
| 9 | GPT-4.1 (2025-04-14) | 94 | #257 |
| 10 | GPT-4.5 | 94 | #278 |
| 11 | Gemini 2.5 Pro (Preview 03-25) | 93.33 | #126 |
| 12 | GPT-4 Turbo | 93.33 | #417 |
| 13 | O3 (2025-04-16) | 93.33 | #125 |
| 14 | GPT-4o (2024-11-20) | 92.67 | #392 |
| 15 | GPT-4.1 Mini | 92 | #363 |
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-reasoning-jigsaw · How It Works · Data refreshed daily, snapshot 2026-11-05.