HELM VHELM - Blink Knowledge - Forensic Detection: leaderboard
Metric: Prefix Quasi-Exact Match (%). Source: crfm.stanford.edu. 48 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.5 | 97.73 |
| 2 | GPT-4.1 Mini | 96.97 |
| 3 | O1 (2024-12-17) | 96.97 |
| 4 | Gemini 2.5 Pro (Preview 03-25) | 93.18 |
| 5 | GPT-4.1 (2025-04-14) | 91.67 |
| 6 | GPT-4o (2024-05-13) | 91.67 |
| 7 | O4 Mini (2025-04-16) | 90.15 |
| 8 | Gemini 2.0 Flash (Preview) | 88.64 |
| 9 | O3 (2025-04-16) | 88.64 |
| 10 | GPT-4o (2024-11-20) | 87.88 |
| 11 | Gemini 2.0 Flash | 86.36 |
| 12 | GPT-4o (2024-08-06) | 86.36 |
| 13 | Gemini 2.0 Pro (Preview 02-05) | 86.36 |
| 14 | Gemini 1.5 Pro (002) | 81.82 |
| 15 | Llama 4 Maverick Instruct FP8 | 76.52 |
Interactive version: theaggregate.ai/benchmark?slug=helm-vhelm-blink-knowledge-forensic-detection · How It Works · Data refreshed daily, snapshot 2026-09-19.