MOSS

MOSS appears in Vol.114 AI的2025和DeepSeek们的未来 | 对谈复旦张奇教授 as the large-model project associated with [[FudanUniversity|复旦大学]] and [[ZhangQi|张奇]]. The source does not profile MOSS technically; it uses Zhang’s role as MOSS lead to establish his credibility in natural-language processing and large-model research.

The wiki should therefore treat MOSS as institutional context for the episode’s claims rather than as an independently evaluated model case. Those claims concern DeepSeek’s cost and engineering implications, the LLM Statistical Boundary of current models, and the difficulty of Model Post-Training Bottleneck work after pretraining.

Connections

• [[ZhangQi|张奇]] and [[FudanUniversity|复旦大学]] — person and institution attached to MOSS in the source.
• DeepSeek and Open Source AI Models — adjacent Chinese large-model context.
• LLM Statistical Boundary, Frontier Model Scaling, and Model Post-Training Bottleneck — model-progress and limitation concepts tied to the discussion.