STACKQUADRANT

zhihu/ZhiLight

Inference Engines

A highly optimized LLM inference acceleration engine for Llama and its variants.

5.1
GitHub Metrics
Stars
908
Forks
104
Open Issues
6
Watchers
52
Contributors
9
Weekly Commits
0
Language
C++
License
Apache-2.0
Last Commit
Mar 18, 2026
Created
Dec 6, 2024
Latest Release
v0.4.8
Release Date
Dec 10, 2024
Synced: Aug 28, 2026
Quality Scores
Documentation Qualityw: 20%
4.1

No dedicated docs site. Description: 80 chars. Stars signal: 908. Contributors: 9. Score: 4.1/10

Community Healthw: 20%
5.5

Stars: 908. Contributors: 9. Watchers: 52. Forks: 104. Issue ratio: 0.7%. Score: 5.5/10

Maintenance Velocityw: 15%
3.3

Last commit: 163d ago. Weekly commits: 0. Latest release: v0.4.8. Maturity bonus: 1.7y old. Score: 3.3/10

API Design & DXw: 20%
7.0

Stars/issues ratio: 151. No dedicated API docs. Permissive license: Apache-2.0. Popularity signal: 908 stars. Score: 7/10

Production Readinessw: 15%
4.3

Battle-tested: 908 stars. Peer review: 9 contributors. Versioned: v0.4.8. Licensed: Apache-2.0. Age: 1.7 years. Maintenance: last commit 163d ago. Score: 4.3/10

Ecosystem Integrationw: 10%
6.1

Fork interest: 104. Ecosystem: C++. Integration-friendly: Apache-2.0. Adoption: 908 stars. Score: 6.1/10

Tags
cudadeepseek-r1gptinference-enginellamallmllm-inferencellm-servingmodel-servingpytorch
Radar
Documentation Quality
Community Health
Maintenance Velocity
API Design & DX
Production Readiness
Ecosystem Integration