Open Chinese LLM - CMMLU — leaderboard

Metric: Accuracy (%). Source: huggingface.co. 338 models tracked.

Top models

#ModelScore
1Qwen 2 72B78.66
2Qwen 2 72B Instruct78.4
3Higgs-Llama-3-70B75.33
4free-evo-qwen72B-v0.8-re73.88
5TW3-JRGL-v273.73
6Llama 3 70B Synthia V3.573.46
7Rhea-72B-v0.573.31
8Meta-LLama-3-Cat-Smaug-LLama-70B72.48
9autotrain-llama3-70B-orpo-v172.42
10dolphin-2.9.1-llama-3-70B72.34
11Smaug-72B-v0.172.32
12Llama 3 70B Japanese Suzume Vector V0.172.14
13Llama3-TenyxChat-70B72.02
14Le_Triomphant-ECE-TW371.86
1534B-beta71.71

Interactive version: theaggregate.ai/benchmark?slug=open-chinese-llm-cmmlu · How the rankings work · Data refreshed daily, snapshot 2026-07-22.