Unverified50% confidenceBenchmarkExact time
在RTX 5090上，Muse Glimmer生成速度从每秒75个Token提升到233个Token，提速约3.1倍
1
Sources
50%
Confidence
Long-term
Relevance
8/15/2026
First Seen
Sources
Related Claims
UnverifiedRTX 4060生成速度可达GTX 1060的3-5倍79% similarUnverifiedGemma量化模型在高性能笔记本电脑上Token生成速度可达每秒20-50个，纯CPU推理约5-15 tokens/秒，INT4量化下NVIDIA RTX 4060/4070可将推理速度提升3-5倍75% similarUnverifiedComposer 2.5的Token生成速度达到每秒200个甚至更高70% similarUnverified千问3.6 27B FP8模型开启MTP后正常速度约38 token/s，平均测速可达70 token/s以上70% similarUnverified实测在RTX 3060（6GB显存）、32GB RAM上运行预量化Krea2 Turbo，1024×1024分辨率8步Euler采样，速度为1.78 s/it，总耗时17.64秒70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/754283API
curl https://kongchang.com/api/v1/knowledge/claims/754283MCP
get_claim(id=754283)