Kimi CLI is your next CLI agent.

A scrollable-tiling Wayland compositor.

Fast and memory-efficient exact attention

llvm-mctoll

A fast communication-overlapping library for tensor/expert parallelism on GPUs.

FP16xINT4 LLM inference kernel that can achieve near-ideal ~4x speedups up to medium batchsizes of 16-32 tokens.

Async RL Training at Scale

A collection of LLM memes

Learn CUDA with PyTorch

Renderer for the harmony response format to be used with gpt-oss

Tensor library & inference framework for machine learning

A CLI tool for managing Claude instances with git worktree

CUDA kernel author's tools

A Distributed Attention Towards Linear Scalability for Ultra-Long Context, Heterogeneous Data Training

Simple & Scalable Pretraining for Neural Architecture Research

Fused Qwen3 MoE layer for faster training, compatible with HF Transformers, LoRA, 4-bit quant, Unsloth

A Tree Search Library with Flexible API for LLM Inference-Time Scaling

MiniHF is an inference, human preference data collection, and fine-tuning tool for local language models. It is intended to help the user develop their prompts into full models.

PyTorch Single Controller

slime is an LLM post-training framework for RL Scaling.

Mirage Persistent Kernel: Compiling LLMs into a MegaKernel

Manage resources and move them between hardware contexts

🚀 Efficient implementations of state-of-the-art linear attention models

Tenstorrent Blackhole P100/P150 card RISC-V Linux demo 🐧

Fast low-bit matmul kernels in Triton

A collection of formalized statements of conjectures in Lean.