Kaggle MultiLoKo - Average v2 (Tasks v6, German v3, Korean and Turkish v7): leaderboard

Native model scores in the source’s stated configuration. Metric: Exact Match (%). Source: https://www.kaggle.com/benchmarks/metaresearch/multiloko

Metric: Exact Match (%). Source: www.kaggle.com. Saturation forecast: Around June 2028. 16 models tracked. Updated 17 Nov 2026.

Top models

#ModelScoreOverall rank
1Gemini 3.1 Pro (Preview)57.03#99
2Gemini 3.8 Flash53.23#39
3GPT-6.1 Sol52.85#12
4GPT-6 Astra52.81#8
5Gemini 4 Argon47.57#14
6GPT-6 Luna40.65#105
7Grok 4.736.76#48
8Qwen 3.8 Max35.95#64
9Claude Opus 5.532.93#7
10GLM-531.3#267
11Kimi K328#59
12Claude Sonnet 5.526.67#21
13Claude Opus 4.822.44#62
14GLM-5.3 Flash22.27#136
15Claude Sonnet 522.01#118

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=kaggle-multiloko-average-v2-tasks-v6-german-v3-korean-and-turkish-v7 · How It Works · Data refreshed daily, snapshot 2026-11-17.