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

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

SGLang is a fast serving framework for large language models and vision language models.

Run any open-source LLMs, such as DeepSeek and Llama, as OpenAI compatible API endpoint in the cloud.

SkyPilot: Run AI and batch jobs on any infra (Kubernetes or 16+ clouds). Get unified execution, cost savings, and high GPU availability via a simple interface.

The easiest way to serve AI apps and models - Build Model Inference APIs, Job queues, LLM apps, Multi-model pipelines, and more!

Superduper: End-to-end framework for building custom AI applications and agents.

Multi-LoRA inference server that scales to 1000s of fine-tuned LLMs

MoBA: Mixture of Block Attention for Long-Context LLMs

High-performance inference framework for large language models, focusing on efficiency, flexibility, and availability.

A high-performance ML model serving framework, offers dynamic batching and CPU/GPU pipelines to fully exploit your compute machine

Community maintained hardware plugin for vLLM on Ascend

Efficient AI Inference & Serving

A ChatGPT(GPT-3.5) & GPT-4 Workload Trace to Optimize LLM Serving Systems

A tiny yet powerful LLM inference system tailored for researching purpose. vLLM-equivalent performance with only 2k lines of code (2% of vLLM).

🪶 Lightweight OpenAI drop-in replacement for Kubernetes

From Hours to Minutes: Lossless Acceleration of Ultra Long Sequence Generation

A simple service that integrates vLLM with Ray Serve for fast and scalable LLM serving.

Repo for "Z1: Efficient Test-time Scaling with Code"

Friendli: the fastest serving engine for generative AI