HELM VHELM - Blink Perception - Semantic Correspondence: leaderboard
Metric: Prefix Quasi-Exact Match (%). Source: crfm.stanford.edu. 48 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O3 (2025-04-16) | 83.45 |
| 2 | O1 (2024-12-17) | 78.42 |
| 3 | O4 Mini (2025-04-16) | 77.7 |
| 4 | GPT-4.5 | 76.98 |
| 5 | GPT-4.1 (2025-04-14) | 74.1 |
| 6 | Gemini 2.0 Flash Lite | 74.1 |
| 7 | Gemini 2.5 Pro (Preview 03-25) | 73.38 |
| 8 | Gemini 2.0 Flash (Preview) | 73.38 |
| 9 | Gemini 1.5 Pro (002) | 71.22 |
| 10 | Gemini 2.0 Pro (Preview 02-05) | 71.22 |
| 11 | Gemini 2.0 Flash | 68.35 |
| 12 | GPT-4o (2024-08-06) | 68.35 |
| 13 | GPT-4o (2024-05-13) | 66.91 |
| 14 | Llama 4 Maverick Instruct FP8 | 66.91 |
| 15 | Claude 3.5 Sonnet (20240620) | 65.47 |
Interactive version: theaggregate.ai/benchmark?slug=helm-vhelm-blink-perception-semantic-correspondence · How It Works · Data refreshed daily, snapshot 2026-09-19.