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

Multilingual Math Reasoning on MT-AIME

85.67Mean@3

DeepSeek-V3.2

Updated 1mo ago

Evaluation Results

Method	Links
DeepSeek-V3.2 2026.04		85.67
o4-mini-high 2026.04		75.33
DeepSeek-V3.2 2026.04		53
x1-Qwen3-32B 2026.04		34.5
Qwen3-32B 2026.04		33.89
x1-Qwen3-14B 2026.04		33.11
Qwen3-14B 2026.04		29.22
x1-DeepSeek-R1-Distill-Qwen-7B 2026.04		27
DeepSeek-R1-Distill-Qwen-7B 2026.04		25.83
x1-Qwen3-4B 2026.04		22.83
x1-Qwen3-32B 2026.04		22.11
Qwen3-32B 2026.04		21.83
Qwen3-4B 2026.04		21.78
x1-Qwen3-14B 2026.04		19.44
Qwen3-14B 2026.04		19.33
x1-DeepSeek-R1-Distill-Llama-8B 2026.04		17
DeepSeek-R1-Distill-Llama-8B 2026.04		14.44
x1-Qwen3-4B 2026.04		13.56
Qwen3-4B 2026.04		12.89
x1-DeepSeek-R1-Distill-Qwen-7B 2026.04		9
DeepSeek-R1-Distill-Qwen-7B 2026.04		8.33
x1-DeepSeek-R1-Distill-Llama-8B 2026.04		2.89
DeepSeek-R1-Distill-Llama-8B 2026.04		2.67