CUDA Tile IR is an MLIR-based intermediate representation and compiler infrastructure for CUDA kernel optimization, focusing on tile-based computation patterns and optimizations targeting NVIDIA te…

An early research stage expert-parallel load balancer for MoE models based on linear programming.

MiMo-V2-Flash: Efficient Reasoning, Coding, and Agentic Foundation Model

cuTile is a programming model for writing parallel kernels for NVIDIA GPUs

Tile-Based Runtime for Ultra-Low-Latency LLM Inference

DeepSeek-V3.2-Exp DSA Warmup Lightning Indexer training operator based on tilelang

WaferLLM: Large Language Model Inference at Wafer Scale

ReDesign of https://github.com/microsoft/BitBLAS

"AI-Trader: Can AI Beat the Market?" Live Trading Bench: https://ai4trade.ai Tech Report Link: https://arxiv.org/abs/2512.10971

RLinf is a flexible and scalable open-source infrastructure designed for post-training foundation models (LLMs, VLMs, VLAs) via reinforcement learning.

Building the Virtuous Cycle for AI-driven LLM Systems

Open Source Continuous Inference Benchmarking - GB200 NVL72 vs MI355X vs B200 vs H200 vs MI325X & soon™ TPUv6e/v7/Trainium2/3/GB300 NVL72 - DeepSeek 670B MoE, GPTOSS

a size profiler for cuda binary

Ascend TileLang adapter

Open ABI and FFI for Machine Learning Systems

NVIDIA NVSHMEM is a parallel programming interface for NVIDIA GPUs based on OpenSHMEM. NVSHMEM can significantly reduce multi-process communication and coordination overheads by allowing programmer…

A verification tool for ensuring parallelization equivalence in distributed model training.

FlashInfer: Kernel Library for LLM Serving

Open-Source Frontier Voice AI

gpt-oss-120b and gpt-oss-20b are two open-weight language models by OpenAI

Tile primitives for speedy kernels

Tile-based language built for AI computation across all scales

Kimi K2 is the large language model series developed by Moonshot AI team