Expired60% confidenceOpinionExact time
对于 198B 总参数的稀疏 MoE 模型，实际激活参数量可能仅为 20-40B 量级
1
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
5/29/2026
First Seen
Valid until: 8/27/2026(expired)
Sources
Step 3.7 Flash: Deep Dive into the 198B Sparse MoE Multimodal Model
twitterlmsysorg5/29/2026
Related Entities
Related Claims
UnverifiedTiny-MoE模型总参数量约200M，但每个token激活的参数仅约90M83% similarUnverifiedMoE模型中'35B'指总参数量，每个token只激活少量Expert（通常2-4个），实际计算量可能仅相当于6-7B的Dense模型81% similarUnverified千问三235B MoE总参数量2350亿，但每次推理实际激活的参数约22B75% similarUnverifiedAnalysis speculates that the model is an MoE model with 2 to 4 trillion parameters, activating only approximately 150 billion parameters per token.73% similarUnverifiedMoE模型采用稀疏激活，Qwen 3.5 122B总参数122B但每次推理仅激活约10B，需将全部参数加载进内存，对内存容量极度敏感70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/607API
curl https://kongchang.com/api/v1/knowledge/claims/607MCP
get_claim(id=607)