HELM VHELM - A Okvqa Chinese: leaderboard
Metric: PEM (Translate). Source: crfm.stanford.edu. 48 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro (Preview 03-25) | 89.3 |
| 2 | O3 (2025-04-16) | 86.5 |
| 3 | GPT-4.5 | 86.4 |
| 4 | GPT-4.1 (2025-04-14) | 85.4 |
| 5 | Gemini 2.0 Pro (Preview 02-05) | 84.4 |
| 6 | GPT-4o (2024-11-20) | 84.3 |
| 7 | Gemini 2.0 Flash (Preview) | 84.1 |
| 8 | Gemini 2.0 Flash | 84.1 |
| 9 | GPT-4o (2024-05-13) | 84.1 |
| 10 | O4 Mini (2025-04-16) | 83.9 |
| 11 | GPT-4o (2024-08-06) | 83.7 |
| 12 | Gemini 1.5 Pro (002) | 83.1 |
| 13 | Gemini 2.0 Flash Lite (Preview 02-05) | 83 |
| 14 | Gemini 2.0 Flash Lite | 82.9 |
| 15 | GPT-4.1 Mini | 82.3 |
Interactive version: theaggregate.ai/benchmark?slug=helm-vhelm-a-okvqa-chinese · How It Works · Data refreshed daily, snapshot 2026-09-08.