Kaggle MultiLoKo - Traditional Mandarin: leaderboard

Metric: Exact Match (%). Source: www.kaggle.com. 48 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview)75.4
2Gemini 3.5 Flash66.8
3Gemini 3 Flash (Preview)66.8
4Gemini 2.5 Pro (Preview 06-05)57.6
5Gemini 3.1 Flash Lite (Preview)56.8
6DeepSeek V347.2
7DeepSeek R1 052847.2
8Gemini 2.0 Flash46.8
9Gemini 2.5 Flash (Preview 05-20)46.4
10GPT-5.445.2
11O143.6
12GPT-543.2
13Claude Opus 442.8
14Claude Opus 4.141.6
15Gemini 1.5 Pro (002)40.8

Interactive version: theaggregate.ai/benchmark?slug=kaggle-multiloko-traditional-mandarin · How It Works · Data refreshed daily, snapshot 2026-09-19.