Unverified50% confidenceFactExact time
At small scale, the FLOPs share of the MLP layer is approximately 44%, rising to approximately 80% when scaled to 175B parameters
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
Unverified业界顶级集群的MFU（模型算力利用率）通常在35%-60%之间68% similarUnverified单条推理通常只能占用GPU SM利用率的5%～15%，而合理的批次大小可将利用率提升至60%～90%67% similarUnverifiedIn MoE models, dense layer parameters typically account for less than 20% of total parameters, but every token must pass through them, so they require higher-precision protection.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/904275API
curl https://kongchang.com/api/v1/knowledge/claims/904275MCP
get_claim(id=904275)