List your company

Language Modeling on Pile (test)

59.4AccuracyYuan3.0-1T Base
53.99255.39656.858.204Jan 20, 2026

Evaluation results

MethodLinks
Yuan3.0-1T Base#Shots=-, Architecture=MoE, # activated params=68.5B, # total params=1010B2026.0159.4
DeepSeek-V3-Base#Shots=-, Architecture=MoE, # activated params=37B, # total params=671B2026.0154.8
LLaMA-3.1-405B Base#Shots=-, Architecture=Dense, # activated params=405B, # total params=405B2026.0154.2