Easy, Fast, and Scalable Multimodal AI

verl: Volcano Engine Reinforcement Learning for LLMs

TPU inference for vLLM, with unified JAX and PyTorch support.

SkyRL: A Modular Full-stack RL Library for LLMs

Post-training with Tinker

[NeurIPS 2025] Scaling Speculative Decoding with Lookahead Reasoning

Common recipes to run vLLM

Open-source implementation of AlphaEvolve

Nano vLLM

Achieve state of the art inference performance with modern accelerators on Kubernetes

A Datacenter Scale Distributed Inference Serving Framework

ArcticInference: vLLM plugin for high-throughput, low-latency inference

MiMo: Unlocking the Reasoning Potential of Language Model – From Pretraining to Posttraining

Democratizing Reinforcement Learning for LLMs

[ACL 2025 Long Main] Language Model Fine-Tuning on Scaled Survey Data for Predicting Distributions of Public Opinions

NumPy aware dynamic Python compiler using LLVM

[NeurIPS 2025] Simple extension on vLLM to help you speed up reasoning model without training.

A collection of GPT system prompts and various prompt injection/leaking knowledge.

FAIR Sequence Modeling Toolkit 2

Mooncake is the serving platform for Kimi, a leading LLM service provided by Moonshot AI.

Fast, Flexible and Portable Structured Generation

Helpful tools and examples for working with flex-attention

Efficient Triton Kernels for LLM Training

A throughput-oriented high-performance serving framework for LLMs

Enabling PyTorch on XLA Devices (e.g. Google TPU)

A minimal implementation of vllm.

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

PyTorch native quantization and sparsity for training and inference