DeepSeek 4 Flash local inference engine for Metal and CUDA

TokenSpeed is a speed-of-light LLM inference engine.

cudnn_frontend provides a c++ wrapper for the cudnn backend API and samples on how to use it

The fastest priority queue

Get the main content of any page as Markdown.

A kernel library written in tilelang

The Patterns of Scalable, Reliable, and Performant Large-Scale Systems

sitp: run nanochat by building teenygrad from scratch: the bridge from micrograd to tinygrad!

The open-source agent-serving project

An Extensible Deep Learning Library

The agent that grows with you

Fast state-of-the-art image and video segmentation in portable C/C++

A nearly complete collection of prefix sum algorithms implemented in CUDA, D3D12, Unity and WGPU. Theoretically portable to all wave/warp/subgroup sizes.

A modern causal profiler built leveraging Linux tracepoints

VisuTwin Canvas

Protocol Buffers implementation in C

8x Faster JavaScript 3D Library.

A simple HTTP server written from scratch as a teaching tool to teach Unix network program architectures

Build your own high performance LLM inference engine in C++ and CUDA - a smaller version of vLLM

A small, portable and extensible C++ 3D coding framework

Distribute and run LLMs with a single file.

📚LeetCUDA: Modern CUDA Learn Notes with PyTorch for Beginners🐑, 200+ CUDA Kernels, Tensor Cores, HGEMM, FA-2 MMA.🎉

Fast, accurate & comprehensive text measurement & layout

Building the Virtuous Cycle for AI-driven LLM Systems

Machine learning framework written in C.

GLyphy is an implementation of the Slug algorithm for GPU text rasterization

AI Tensor Engine for ROCm

WebGPU implementation of Eric Lengyel's Slug algorithm for resolution-independent vector text rendering on the GPU