🦜🔗 The platform for reliable agents.

A framework for efficient model inference with omni-modality models

⭐AI-driven public opinion & trend monitor with multi-platform aggregation, RSS, and smart alerts.🎯 告别信息过载，你的 AI 舆情监控助手与热点筛选工具！聚合多平台热点 + RSS 订阅，支持关键词精准筛选。AI 翻译 + AI 分析简报直推手机，也支持接入 MCP 架构，赋能 AI 自然语言对…

UniVideo: Unified Understanding, Generation, and Editing for Videos

RLinf: Reinforcement Learning Infrastructure for Embodied and Agentic AI

Spirit-v1.5: A Robotic Foundation Model by Spirit AI

The open source coding agent.

Ray is an AI compute engine. Ray consists of a core distributed runtime and a set of AI Libraries for accelerating ML workloads.

Ongoing research training transformer models at scale

Official Python inference and LoRA trainer package for the LTX-2 audio–video generative model.

RL research on Android devices.

A Cross-Platform Backend for High-Performance Sparse Convolutions

Echo-4o

Vision Bridge Transformer at Scale

The official implementation for [NeurIPS2025 Oral] Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free

Official inference repo for FLUX.2 models

Kandinsky 5.0: A family of diffusion models for Video & Image generation

Momentum Human Rig is an anatomically-inspired parametric full-body digital human model developed at Meta. It includes: A parametric body skeletal model; A realistic 3D mesh skinned to the skeleton…

PyTorch implementation of JiT https://arxiv.org/abs/2511.13720

SAM 3D Objects

VLAC: A Vision-Language-Action-Critic Model for Robotic Real-World Reinforcement Learning

Open-source framework for the research and development of foundation models.

Open-source Trading OS with pluggable AI brain | From market data → AI reasoning → Trade execution | Self-hosted & Multi-exchange

Autonomous GPU Kernel Generation via Deep Agents

Testing adaptation of the DINOv2/3 encoders for vision tasks with Low-Rank Adaptation (LoRA)

微舆：人人可用的多Agent舆情分析助手，打破信息茧房，还原舆情原貌，预测未来走向，辅助决策！从0实现，不依赖任何框架。

A simple, unified multimodal models training engine. Lean, flexible, and built for hacking at scale.

EasyR1: An Efficient, Scalable, Multi-Modality RL Training Framework based on veRL