Code at the speed of thought – Zed is a high-performance, multiplayer code editor from the creators of Atom and Tree-sitter.

OpenSpiel is a collection of environments and algorithms for research in general reinforcement learning and search/planning in games.

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

VibeRL is a Reinforcement Learning framework built essentially through vibe coding with Kimi K2.

Use Kimi latest model(kimi-k2-0711-preview) to drive your Claude Code.

Code release for paper "Test-Time Training Done Right"

Triton implementation of FlashAttention2 that adds Custom Masks.

FlashMLA: Efficient Multi-head Latent Attention Kernels

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

Open-source unified multimodal model

Kimi-Audio, an open-source audio foundation model excelling in audio understanding, generation, and conversation

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

Kimi-VL: Mixture-of-Experts Vision-Language Model for Multimodal Reasoning, Long-Context Understanding, and Strong Agent Capabilities

Converts KaTeX escape characters with MathJax escape characters

NanoGPT (124M) in 3 minutes

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

Analyze computation-communication overlap in V3/R1.

Efficient Triton Kernels for LLM Training

🐳 Efficient Triton implementations for "Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention"

[MLSys'25] QServe: W4A8KV4 Quantization and System Co-design for Efficient LLM Serving; [MLSys'25] LServe: Efficient Long-sequence LLM Serving with Unified Sparse Attention

My learning notes/codes for ML SYS.

Concurrent Python made simple

verl: Volcano Engine Reinforcement Learning for LLMs

A delightfully minimal, yet remarkably powerful AI Shell Assistant.

GTP engine and self-play learning in Go

Portrait4D: Learning One-Shot 4D Head Avatar Synthesis using Synthetic Data (CVPR 24); Portrait4D-v2: Pseudo Multi-View Data Creates Better 4D Head Synthesizer (ECCV 2024)

High-resolution models for human tasks.

💥 Blazing fast terminal file manager written in Rust, based on async I/O.