Unverified70% confidenceFactExact time
In the Transformer architecture, the computational cost per generated token is approximately proportional to the number of activated parameters; for dense models the activated parameter count equals the total parameter count, while for MoE models the activated parameter count is far smaller than the total parameter count.
2
Sources
70%
Confidence
Long-term
Relevance
8/26/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn the Transformer architecture, there is an approximately linear relationship between model parameter count and inference latency.79% similarUnverified以Transformer架构为基础的语言模型，其推理计算量与参数规模大致呈正比，参数越多单次推理所需的FLOPs越高78% similarVerifiedMainstream generative large models are based on the Transformer architecture and have a probabilistic selection mechanism during output, which means the same question may produce different answers.78% similarUnverified在Transformer自回归推理模式下，输出Token数量与计算开销之间存在超线性关系75% similarUnverified在Transformer架构中，对于拥有N个参数的模型，生成每个token大约需要2N次浮点运算75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/802774API
curl https://kongchang.com/api/v1/knowledge/claims/802774MCP
get_claim(id=802774)