Apache-2.0freeRuns locallyDev & Coding

vLLM: Open-Source LLM Serving Engine

Apache-2.0 high-throughput LLM inference and serving engine (PagedAttention).

Official vLLM v1 process architecture diagram from the repository docs — Official project image. Source · License: Apache-2.0.

License: Apache-2.0
Stars: 94,459 (2026-10-21)
Platforms: Linux, NVIDIA, AMD, CPU, Python
First seen: 2023-02-09

TL;DR. vLLM is Apache-2.0 high-throughput LLM inference and serving engine (PagedAttention). License: Apache-2.0 (verified). Platforms: Linux, NVIDIA, AMD, CPU, Python. Price: free. Runs locally: yes. It is an alternative to Closed LLM APIs, ChatGPT.

Name	vLLM
Language	Python
License	Apache-2.0
Pricing	free
Platforms	Linux, NVIDIA, AMD, CPU, Python
Runs locally	Yes
Stats	94,459 stars · 23,675 forks (as of 2026-10-21)
First seen	2023-02-09
Repository	https://github.com/vllm-project/vllm

vLLM is an inference and serving engine for large language models. Official hosts we opened on 2026-10-21: vllm-project/vllm, LICENSE, release v0.32.0, docs.vllm.ai, and vllm.ai. GitHub API that day: 94,459 stars, 23,675 forks, created 2023-02-09T11:23:20Z, last push 2026-10-21T02:04:25Z, language Python, homepage vllm.ai. Latest release v0.32.0, published 2026-10-18T06:50:48Z. LICENSE is Apache-2.0.

The README tagline: “A high-throughput and memory-efficient inference and serving engine for LLMs.” It names PagedAttention for KV memory, continuous batching, an OpenAI-compatible API server, and “200+ model architectures on Hugging Face.” Throughput adjectives stay vendor.

It hangs off closed LLM APIs and ChatGPT alternatives as a serving stack. Compare DeepInfer vs vLLM. Self-hosted UIs: Open WebUI, LibreChat, Jan. Models on this site that name vLLM in official cards include Ornith-1.5-122B (card recipe vllm ≥ 0.19.1, 2026-10-22), DeepSeek-V4.1-Flash, Hy-Memory, Hy-SimulMT (card deploy path vLLM >=0.31), Ling-3.1-flash, Muse Spark Open, and Nemotron 3.5 Super VL. Ornith vs Flash: Ornith-1.5-122B vs DeepSeek-V4.1-Flash.

Key facts (2026-10-21)

Field	Value	Source
Stars / forks	94,459 / 23,675	GitHub API
Created	2023-02-09T11:23:20Z	same
Last push	2026-10-21T02:04:25Z	same
License	Apache-2.0	`LICENSE`
Language	Python	GitHub
Latest release	v0.32.0, 2026-10-18T06:50:48Z	GitHub Releases
Docs / site	docs.vllm.ai / vllm.ai	README

README hardware labels (copied): NVIDIA GPUs, AMD GPUs, Intel GPUs, x86/ARM/PowerPC CPUs, plus plugins named for Google TPUs, Intel Gaudi, IBM Spyre, Huawei Ascend, Rebellions NPU, Apple Silicon, MetaX GPU. We did not open each plugin’s hardware matrix.

How to run (from the README, not tested)

The README’s install sketch:

uv pip install vllm
vllm download-kernels

Docs: installation and quickstart. forkalt has not run these commands.

DeepInfer is DeepSeek’s Apache-2.0 engine whose README credits vLLM as the primary original base and lists DeepSeek-V4.1-Flash only. SGLang is named on several model cards here (37,519 stars, Apache-2.0, v0.5.22 in the 2026-10-21 check); this site does not yet have a dedicated SGLang tool page.

What we did not invent

Serving throughput, VRAM, and “fast/cheap” claims stay vendor.
We do not rank vLLM against DeepInfer or SGLang.

Comparisons

DeepInfer vs vLLM: Which Should You Use?

FAQ

Is vLLM open source?

Yes. LICENSE is Apache-2.0 (checked 2026-10-21). GitHub reports Apache-2.0. Latest release in this check is v0.32.0 (published 2026-10-18). Not legal advice.

How is it different from DeepInfer?

vLLM is a general serving engine (200+ architectures named on the README). DeepInfer is DeepSeek’s framework whose README lists DeepSeek-V4.1-Flash and credits vLLM as the primary original base. See DeepInfer vs vLLM. No speed ranking.

Does it include a chat UI?

No. It is a library plus an OpenAI-compatible API server (README also names Anthropic Messages API and gRPC). Pair it with a UI such as Open WebUI or LibreChat.

Sources & verification

Last verified 2026-10-21 · Last updated 2026-10-21· Edited by forkalt editors · Editorial policy

License notes are not legal advice. Read the original LICENSE. Trademark & disclaimer.