GitHub intelligence

ztxz16/fastllm

fastllm是后端无依赖的高性能大模型推理库。同时支持张量并行推理稠密模型和混合模式推理MOE模型,任意10G以上显卡即可推理满血DeepSeek。双路9004/9005服务器+单显卡部署DeepSeek满血满精度原版模型,单并发20tps;INT4量化模型单并发30tps,多并发可达60+。

Observed OCT 8, 2026 · 00:08 UTC

Facts

Language
C++
License
apache-2.0
Created
MAY 13, 2023
Last push
Last release
None recorded

Measurements

Stars
5,090
Forks
501
7D change
Collectingfrom OCT 9
30D change
Collectingfrom NOV 1
90D change
Collectingfrom DEC 31
Velocity
Collectingfrom OCT 16
Acceleration
Collectingfrom OCT 23
Trajectory
Collectingfrom OCT 16
Trending percentile
Collectingfrom OCT 16

Categories

Tentative categories were assigned with moderate confidence (0.5 to 0.75): they are shown, but they do not count toward a category’s momentum. How categories work

Technologies

Not linked to any curated technology.

Star history

Daily history began OCT 2. The chart appears after 7 days. 1 observed day shown.

5.1K5.1KJUL 10JUL 31AUG 21SEP 11OCT 2OCT 2: 5,090 stars
Observed days (1)
DayStarsForks
OCT 25,090501