moe

A high-throughput and memory-efficient inference and serving engine for LLMs

Unified Efficient Fine-Tuning of 100+ LLMs & VLMs (ACL 2024)

SGLang is a high-performance serving framework for large language models and multimodal models.

Use PEFT or Full-parameter to CPT/SFT/DPO/GRPO 600+ LLMs (Qwen3.6, DeepSeek-V4, GLM-5.1, InternLM3, Llama4, ...) and 300+ MLLMs (Qwen3-VL, Qwen3-Omni, InternVL3.5, Ovis2.5, GLM4.5v, Gemma4, Llava, Phi4, ...) (AAAI 2025).

TensorRT LLM provides users with an easy-to-use Python API to define Large Language Models (LLMs) and supports state-of-the-art optimizations to perform inference efficiently on NVIDIA GPUs. TensorRT LLM also contains components to create Python and C++ runtimes that orchestrate the inference execution in a performant way.

FreeToken brings datacenter-scale model serving to your desktop. Run massive models locally, fast and efficiently.

A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.

FlashInfer: Kernel Library for LLM Serving

:electron: An unofficial https://bgm.tv ui first app client for Android and iOS, built with React Native. 一个无广告、以爱好为驱动、不以盈利为目的、专门做 ACG 的类似豆瓣的追番记录，bgm.tv 第三方客户端。为移动端重新设计，内置大量加强的网页端难以实现的功能，且提供了相当的自定义选项。 目前已适配 iOS / Android。

GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models

【TMM 2025🔥】 Mixture-of-Experts for Large Vision-Language Models

MoBA: Mixture of Block Attention for Long-Context LLMs

UCCL is an efficient communication library for GPUs, covering collectives, P2P (e.g., KV cache transfer, RL weight transfer), and EP (e.g., GPU-driven)

PyTorch Re-Implementation of "The Sparsely-Gated Mixture-of-Experts Layer" by Noam Shazeer et al. https://arxiv.org/abs/1701.06538

Tutel MoE: Optimized Mixture-of-Experts Library, Support GptOss/DeepSeek/Kimi-K2/Qwen3 using FP8/NVFP4/MXFP4

⛷️ LLaMA-MoE: Building Mixture-of-Experts from LLaMA with Continual Pre-training (EMNLP 2024)

cuDNN Frontend is NVIDIA's modern, open-source entry point to the cuDNN library and a growing collection of high-performance open-source kernels.

Adan: Adaptive Nesterov Momentum Algorithm for Faster Optimizing Deep Models

An open-source solution for full parameter fine-tuning of DeepSeek-V3/R1 671B, including complete code and scripts from training to inference, as well as some practical experiences and conclusions. (DeepSeek-V3/R1 满血版 671B 全参数微调的开源解决方案，包含从训练到推理的完整代码和脚本，以及实践中积累一些经验和结论。)

⚡ Native MLX Swift LLM inference server for Apple Silicon. OpenAI-compatible API, SSD streaming for 100B+ MoE models, TurboQuant KV cache compression, MACOS + iOS iPhone app.

[CVPR2026]🚀🚀🚀Official code for the paper "YOLO-Master: MOE-Accelerated with Specialized Transformers for Enhanced Real-time Detection." *(YOLO = You Only Look Once)* 🔥🔥🔥

Pure Rust + CUDA LLM inference engine — no PyTorch, OpenAI-compatible, serves Qwen3 to Kimi-K2

中文Mixtral混合专家大模型（Chinese Mixtral MoE LLMs）

Run MoE models bigger than your RAM. Frontier-size MoE on a 12 GB phone, CPU only, lossless, on stock llama.cpp