Expired50% confidenceFactExact time
Pangu 2.0 uses a sparse architecture employing Mixture of Experts (MoE), activating only a subset of parameters per inference rather than all 505 billion
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026(expired)
Sources
Related Claims
Verified混合专家架构（MoE）在推理时仅激活约10%-30%的总参数量，显著降低单次推理的FLOPs和显存占用70% similarUnverifiedMoE架构中，一个标称800亿参数的模型每次推理时可能只激活约200亿参数，VRAM占用大幅降低65% similarUnverifiedM2.5采用混合专家架构（Mixture of Experts，MoE），每次推理只有10B参数真正参与计算，总参数量可能远超10B64% similarVerifiedMoE架构每次推理时通常仅激活总参数的10%-20%62% similarUnverifiedMoE的8选2配置意味着8个专家中每次只激活2个，实际计算量仅为密集模型的约四分之一61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/47218API
curl https://kongchang.com/api/v1/knowledge/claims/47218MCP
get_claim(id=47218)