vllm-project/

vllm-ascend

GitHub Website

vLLM Ascend is a community-maintained hardware plugin enabling vLLM to run on Ascend NPUs, with releases aligned to upstream vLLM versions and documented prerequisites.

2.6kstars

2.0kforks

2.6kissues

Apache-2.0license

2025since

Star historydaily snapshots by VibeCrowd

Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).

Alternatives & relatedmatched by topic overlap

vllm-project/

vllm

▲ 704/wk

A high-throughput and memory-efficient inference and serving engine for LLMs

88k20kPython

huggingface/

transformers

▲ 266/wk

🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

163k34kPython

pathwaycom/

llm-app

▲ 175/wk

Ready-to-run cloud templates for RAG, AI pipelines, and enterprise search with live data. 🐳Docker-friendly.⚡Always in sync with Sharepoint, Google Drive, S3, Kafka, PostgreSQL, real-time data APIs, and more.

59k1.5kJupyter Notebook

BerriAI/

litellm

▲ 595/wk

The fastest, litest AI Gateway. Rust core with Python SDK. Call 100+ LLM APIs in OpenAI (or native) format with cost tracking, guardrails, load balancing, and logging [Bedrock, Azure, OpenAI, Anthropic, OpenAI, VertexAI, vLLM, Nvidia NIM]

56k10kPython

ray-project/

ray

▲ 65/wk

Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.

43k7.9kPython

qdrant/

qdrant

▲ 144/wk

Qdrant - High-performance, massive-scale Vector Database and Vector Search Engine for the next generation of AI. Also available in the cloud https://cloud.qdrant.io/

34k2.6kRust

Comparevllm-ascend vs vllm vllm-ascend vs transformers vllm-ascend vs llm-app

Reviewgenerated from repository data · Aug 5, 2026

What it is

vLLM Ascend is a community maintained hardware plugin for running vLLM on the Ascend NPU. It adheres to a hardware-pluggable interface to decouple Ascend integration from vLLM, supporting Transformer-like models, MoE, Embedding, and Multi-modal LLMs on Ascend.

How it works

The project provides a plugin that integrates Ascend NPU support into vLLM, following the RFC for hardware pluggable architectures. It exposes maintained branches and release tags corresponding to vLLM versions, with a support matrix detailing model and feature compatibility in the documentation.

Getting started

Prerequisites include hardware (Atlas 800I A2/A3 series, Atlas 300I Duo experimental), OS (Linux), software: Python >= 3.10, < 3.13, CANN == 9.0.1, PyTorch == 2.10.0, TorchNPU == 2.10.0.post2, and a vLLM version matching the plugin. The page provides quick-start guidance via QuickStart and Installation docs for specific versions. The repository also lists access to Ascend resources via HiDevLab for development.

Getting started pages referenced:

QuickStart: https://docs.vllm.ai/projects/ascend/en/v0.23.0rc1/quick_start.html
Installation: https://docs.vllm.ai/projects/ascend/en/v0.23.0rc1/installation.html
QuickStart (stable): https://docs.vllm.ai/projects/ascend/en/v0.18.0/quick_start.html
Installation (stable): https://docs.vllm.ai/projects/ascend/en/v0.18.0/installation.html

Recent releases

Latest releases include:

v0.23.0rc1 (2026-07-19) and v0.23.0rc1 (2026-07-20), first release candidate for v0.23.0 line aligned with upstream vLLM v0.23.0.
v0.22.1rc1 (2026-06-30), first release candidate for v0.22.1 line aligned with upstream vLLM.
v0.21.0rc1 (2026-06-16), first release candidate for v0.21.0 line.
v0.20.2rc1 (2026-06-03), first release candidate for v0.20.2 line.
v0.19.1rc1 (2026-04-30), first release candidate for v0.19.1 line.

Traction

The repository has 2556 stars and 1961 forks. Issues open: 2647. (Counts are as provided in the facts.)

License

Apache-2.0 as shown in the LICENSE file.

SharePost on X LinkedIn