[KongchangAI]
Unverified50% confidenceBenchmarkExact time

以70B的BF16模型为例,产生一个token需读入约140GB权重,打满H100的3.35TB/s显存带宽时每token延迟至少42毫秒,理论TPS上限约24 token/秒

1
Sources
50%
Confidence
Long-term
Relevance
9/23/2026
First Seen

Sources

Related Entities

Related Claims

Cite This Claim

Stable URI
https://kongchang.com/claim/942617
API
curl https://kongchang.com/api/v1/knowledge/claims/942617
MCP
get_claim(id=942617)