MT-Bench PL - Roleplay — leaderboard

Metric: Judge Score (0-10). Source: huggingface.co. 50 models tracked.

Top models

#ModelScore
1Gemma 3 27B (IT)9.95
2aya-expanse-32B9.7
3Gemma 3 12B (IT)9.45
4Gemma 3 4B (IT)9.45
5Bielik-11B-v2.1-Instruct9.45
6Mistral Small 3.19.4
7aya-expanse-8B9.25
8Phi-49.2
9Qwen 2 72B Instruct9.2
10Mistral Small 39.05
11Mixtral 8x22B9.05
12Bielik-11B-v2.2-Instruct9.03
13Mixtral 8x7B8.95
14Mistral-Small-Instruct-24098.9
15Llama 3.1 70B Instruct8.8

Interactive version: theaggregate.ai/benchmark?slug=mt-bench-pl-roleplay · How the rankings work · Data refreshed daily, snapshot 2026-07-22.