Share your thoughts, 1 month free Claude Pro on usSee more

Chinese General Knowledge on CMMLU

90.9Accuracy

Kimi-K2 Base

Updated 19d ago

Evaluation Results

Method	Links
Kimi-K2 Base 2026.01		90.9
DeepSeek-V3.2 Exp Base 2026.01		88.9
DeepSeek-V3.1 Base 2026.01		88.8
MiMo-V2-Flash Base 2026.01		87.4
Qwen3.5-9B 2026.04		84.35
Qwen3-14B 2026.04		82.1
Qwen3-8B 2026.04		77.45
XekRung-8B 2026.04		77.05
SecGPT-14B 2026.04		72.35
Llama-3.3-70B-Instruct 2026.04		72.25
DeepSeekV2-Lite 2026.05		60.93
BEAM 2026.05		60.67
BEAM 2026.05		60.42
Top-K Reduced 2026.05		60.4
Top-K Reduced 2026.05		60.39
Top-K Pruning 2026.05		60
AdaMoE 2026.05		58.85
MoE-Dynamic 2026.05		58.43
MoE-Dynamic 2026.05		58.12
BEAM 2026.05		54.48
AdaMoE 2026.05		52.29
Top-K Reduced 2026.05		45.01
Llama-Primus-Reasoning-8B 2026.04		43.15
Top-K Pruning 2026.05		36.91
Llama-3.1-8B-Instruct 2026.04		34.45