vLLM
Open source · Local LLMs · vllm-project/vllm
What it does
vLLM is a server for running language models fast, for many users at once. It is built for GPUs and gives you an OpenAI-compatible API.
Good for: Serving an open model to a team or an app, where speed and cost per request matter.
vLLM on GitHub (opens in a new tab) Project homepage (opens in a new tab)
Visitor votes, one per browser. Your own vote is kept on this device only. How votes work.
The signals we checked
- Licence
- Apache-2.0 (OSI-approved open-source licence)
- Popularity
- Stars: 93,888 (checked 16 Oct 2026)
- Last commit
- 16 Oct 2026
- Open issues and PRs
- 8,887
- Status
- Active (not archived)
- Self-host
- AdvancedNeeds a capable data-centre or gaming GPU and some Linux server know-how.
- GitHub topics
- amd
- blackwell
- cuda
- deepseek
- deepseek-v3
- gpt
- gpt-oss
- inference
- kimi
- llama
More in local llms
- Ollama Trying open models privately on your own computer, and powering other local AI tools.
- Open WebUI A private chat app for you or a small team, in front of local or cloud models.
- llama.cpp Running models on modest hardware, or getting the most speed out of your own machine.
- LocalAI Moving an app that already uses the OpenAI API onto your own hardware.
No affiliate links on this page, and we are not paid by this project. The summary is our own plain-English description; check the repo for current features before you rely on it.
Stars, licences and commit dates come from the GitHub API, checked 16 Oct 2026. "Last commit" is the last push to the repo. Star counts are a rough popularity signal, not a quality score. There are no affiliate links in this section: links go straight to each project.