Stars
Omnigent is an open-source AI agent framework and meta-harness: orchestrate Claude Code, Codex, Cursor, Pi, and custom agents — swap harnesses without rewriting, enforce policies and sandboxing, an…
DeepSeek Harness: Everything is a Plugin.
MAGI-2-preview: Scaling Video Generation Models Efficiently
A scalable, agentic-first, and HuggingFace-native RL framework for research (9k lines).
Uni-Agent is a framework for training long-horizon agents.
Miles is an enterprise-facing reinforcement learning framework for LLM and VLM post-training, forked from and co-evolving with slime.
Build compute kernels and load them from the Hub.
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
An asynchronous streaming data management module for efficient post-training.
A high-performance RL training-inference weight synchronization framework, designed to enable second-level parameter updates from training to inference in RL workflows
Agent Reinforcement Trainer: train multi-step agents for real-world tasks using GRPO. Give your agents on-the-job training. Reinforcement learning for Qwen3.6, GPT-OSS, Llama, and more!
An Efficient and User-Friendly Scaling Library for Reinforcement Learning with Large Language Models
A PyTorch native platform for training generative AI models
Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ languages.
The RL Bridge for LLM-based Agent Applications. Made Simple & Flexible.
Unleashing the Power of Reinforcement Learning for Math and Code Reasoners
Distributed Compiler and Optimized Parallel Kernels
FlashInfer: Kernel Library for LLM Serving
A dict[str, torch.Tensor] that behaves like one tensor, and TensorClass: typed, batched dataclasses that compile.
A high-performance distributed file system designed to address the challenges of AI training and inference workloads.
A bidirectional pipeline parallelism algorithm for computation-communication overlap in DeepSeek V3/R1 training.
DeepGEMM: clean and efficient BLAS kernel library on GPU
DeepEP: an efficient expert-parallel communication library
FlashMLA: Efficient Multi-head Latent Attention Kernels
Official Repo for Open-Reasoner-Zero
A stand-alone implementation of several NumPy dtype extensions used in machine learning.
Solve Visual Understanding with Reinforced VLMs