Language Understanding on CMMLU

90.1Accuracy

Qwen2-72B

Updated 2d ago

Evaluation Results

Method
Qwen2-72B 2024.07	90.1	-
Qwen2-57B-A14B 2024.07	88.5	-
Qwen1.5-110B 2024.07	88.3	-
Yi-1.5-34B 2024.07	84.8	-
Qwen1.5-72B 2024.07	83.5	-
Qwen1.5-32B 2024.07	82.3	-
Qwen3-8B 2026.01	76.8	-
VIST2-8B 2026.01	75.12	-
Qwen3-VL-8B 2026.01	71.35	-
Qwen2-1.5B 2024.07	70.3	-
Qwen3-4B 2026.01	70.25	-
VIST2-4B 2026.01	69.5	-
Llama-3-70B 2024.07	67.2	-
Qwen3-VL-4B 2026.01	65.42	-
Qwen1.5-1.8B 2024.07	57.8	-
Llama-3-SynE 2024.07	57.34	-
Qwen2-0.5B 2024.07	55.1	-
Mixtral-8x22B 2024.07	53.4	-
Llama-3-Chinese-8B 2024.07	51.2	-
Llama-3-8B 2024.07	51.03	-
DeepSeek 7B (Dense) 2024.01	47.2	-
MAmmoTH2-8B 2024.07	45.9	-
Mistral-7B-v0.3 2024.07	43.72	-
DeepSeekMoE 16B 2024.01	42.5	-
DCLM-7B 2024.07	40.89	-
Galactica-6.7B 2024.07	25.53	-
Phi-2 2024.07	24.2	-
LLAMA-7B 2023.08	-	26.8
LLAMA2-7B 2023.08	-	31.8
Baichuan-7B 2023.08	-	44.4
Baichuan2-7B 2023.08	-	57.1
ChatGLM2-6B 2023.08	-	48.8
InternLM-7B 2023.08	-	51.8
Qwen-7B 2023.08	-	62.2
Qwen-VL 2023.08	-	49.5
LLaMA2 SFT 7B 2024.01	-	36.9
DeepSeek Chat 7B 2024.01	-	51.2
DeepSeekMoE Chat 16B 2024.01	-	49.3
DeepSeek 67B (Dense) 2024.01	-	40.6
GShard 137B 2024.01	-	25.4
DeepSeekMoE 145B 2024.01	-	35.9
DeepSeekMoE 142B (Half Activated) 2024.01	-	31.9
LLaMA2 7B 2024.01	-	32.6
DeepSeekMoE 16B 2024.01	-	42.5