Kaggle MultiLoKo - Cantonese: leaderboard
Metric: Exact Match (%). Source: www.kaggle.com. 48 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 55.33 |
| 2 | Gemini 3.5 Flash | 46.18 |
| 3 | Gemini 3 Flash (Preview) | 44.8 |
| 4 | Gemma 4 26B A4B | 44.4 |
| 5 | Gemini 2.5 Pro (Preview 06-05) | 37.2 |
| 6 | Gemini 3.1 Flash Lite (Preview) | 34 |
| 7 | GPT-5 | 30.4 |
| 8 | DeepSeek R1 0528 | 30.4 |
| 9 | O3 | 28.8 |
| 10 | O1 | 28 |
| 11 | Gemini 2.5 Flash (Preview 05-20) | 28 |
| 12 | DeepSeek V3 | 24.4 |
| 13 | Gemini 2.0 Flash | 23.2 |
| 14 | GPT-5.4 | 22.8 |
| 15 | Claude Opus 4 | 22.8 |
Interactive version: theaggregate.ai/benchmark?slug=kaggle-multiloko-cantonese · How It Works · Data refreshed daily, snapshot 2026-09-19.