UCCL is an efficient communication library for GPUs, covering collectives, P2P (e.g., KV cache transfer, RL weight transfer), and EP (e.g., GPU-driven)

An Asynchronous Reinforcement Learning Engine for Omni-Modal Post-Training at Scale

Miles is an enterprise-facing reinforcement learning framework for LLM and VLM post-training, forked from and co-evolving with slime.

turns your codebase into an autoresearch loop — discovers what to measure, instruments the benchmark, then runs tree search with parallel subagents.

Fully autonomous & self-evolving research from idea to paper. Chat an Idea. Get a Paper. 🦞

CUDA kernels for linear attention variants, written in CuTe DSL and CUTLASS C++.

Just a set of claude/codex skills to be 10x Deevs' engineer

Graph-powered code intelligence engine — indexes codebases into a knowledge graph, exposed via MCP tools for AI agents and a CLI for developers.

Complete reverse-engineering documentation of the full logic of Claude Code CLI Node.js version v2.1.84

MiroThinker is a deep research agent optimized for complex research and prediction tasks. Our latest models, MiroThinker-1.7, achieves 74.0 and 75.3 on the BrowseComp and BrowseComp Zh, respectively.

Harbor is a framework for running agent evaluations and creating and using RL environments.

Infiniband Verbs Performance Tests

d9d - d[istribute]d - distributed training framework based on PyTorch that tries to be efficient yet hackable

SANA: Efficient High-Resolution Image Synthesis with Linear Diffusion Transformer

Lightning-Fast, On-Device, Multilingual TTS — running natively via ONNX.

DeepEP: an efficient expert-parallel communication library

SkyRL: A Modular Full-stack RL Library for LLMs

Cursor for Minecraft

🔥 LLM-powered GPU kernel synthesis: Train models to convert PyTorch ops into optimized Triton kernels via SFT+RL. Multi-turn compilation feedback, cross-platform NVIDIA/AMD, Kernelbook + KernelBench

A Datacenter Scale Distributed Inference Serving Framework

[NeurIPS' 25] Benchmark for evaluating TTS models on complex prosodic, expressiveness, and linguistic challenges.

Tile-based language built for AI computation across all scales

Open-source SOTA multi-image editing model

Ongoing research training transformer models at scale

Multi-node distributed LLM training framework

TensorZero is an open-source LLMOps platform that unifies an LLM gateway, observability, evaluation, optimization, and experimentation.

Building AI agents, atomically

Our library for RL environments + evals