Fast Polar Decomposition for Muon

nikki is an experimental translation and runtime layer that implements a restricted, buffer-oriented subset of the Vulkan compute API over WebGPU

DFlash: Block Diffusion for Flash Speculative Decoding

A collection of UI designs implemented in Godot 4

Aurora optimizer release

An asynchronous streaming data management module for efficient post-training.

Mobile and Web client for Codex and Claude Code, with realtime voice, encryption and fully featured

A PyTorch native library for training speculative decoding models

OmX - Oh My codeX: Your codex is not alone. Add hooks, agent teams, HUDs, and so much more.

An agent-managed museum exhibit, built in Rust with Gajae-Code / LazyCodex — developed and maintained with no human intervention.

The AI that really does things. Any OS. Any Platform. The lobster way. 🦞

Checkpoint-engine is a simple middleware to update model weights in LLM inference engines

[NeurIPS 2025 Spotlight] Reasoning Environments for Reinforcement Learning with Verifiable Rewards

A compact implementation of SGLang, designed to demystify the complexities of modern LLM serving systems.

DLBlas: clean and efficient kernels

P1: Mastering Physics Olympiads with Reinforcement Learning

(ICLR 2026) Unveiling Super Experts in Mixture-of-Experts Large Language Models

Dion optimizer algorithm

VeOmni: Scaling Any Modality Model Training with Model-Centric Distributed Recipe Zoo

Physics of Language Models: Part 4.2, Canon Layers at Scale where Synthetic Pretraining Resonates in Reality

A Scientific Multimodal Foundation Model

Pre-trained, Scalable, High-performance Reward Models via Policy Discriminative Learning.

Litex: The Language Where Mathematics Verifies Itself.

Official implement on InternBootCamp

Nano vLLM

An Efficient and User-Friendly Scaling Library for Reinforcement Learning with Large Language Models