Unverified50% confidenceFactExact time
Kimi 2.5每次前向传播只计算约300亿参数，尽管总参数量达1万亿
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
Valid until: 9/30/2026
Sources
Related Claims
Unverified分析推测该模型是2到4万亿参数规模的MoE模型，每个Token仅激活约1500亿参数71% similarUnverified1000亿参数的模型处理单次推理请求需约80GB显存，百亿参数级别模型仅需约20GB69% similarUnverified苹果的端侧模型号称有200亿参数，采用MoE式部分激活机制，每次推理只激活10亿到40亿参数66% similarUnverified若MoE模型每次推理仅激活约1/8的参数，实际计算开销可能仅相当于百亿参数级别的稠密模型65% similarUnverified目前能在手机端流畅运行的模型参数量通常在30亿至70亿级别，与云端数千亿参数的旗舰模型存在数量级差距65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/51431API
curl https://kongchang.com/api/v1/knowledge/claims/51431MCP
get_claim(id=51431)