Private, Offline, Everywhere. We aim to bring the power of Large Language Models (LLMs) to consumer-grade devices—from mobile phones to laptops—without relying on the cloud. By leveraging the linear complexity of the RWKV architecture, we make AI accessible even on low-compute devices.

隐私、离线、无处不在。 我们的目标是将大语言模型（LLM）的能力带入普通用户的手机和笔记本电脑，无需依赖昂贵的云端算力。借助 RWKV 架构的线性复杂度优势，我们让 AI 在低算力设备上也能流畅运行。

Our ecosystem consists of three key layers, ensuring high performance across platforms: 我们的生态系统由三个核心层组成，确保了跨平台的高性能表现：

Repository (仓库)	Role (角色)	Tech Stack (技术栈)
📱 RWKV_APP	[Frontend] The UI/UX layer. Manages model weights, chat interface, and platform-specific logic. [前端交互] 负责 UI 渲染、权重管理及平台特定代码。
🔗 rwkv_mobile_flutter	[Bridge] The communication bridge. Uses Dart FFI to connect the Flutter UI with the C++ backend. [通信桥梁] 通过 Dart FFI 连接 Flutter 前端与 C++ 后端，实现非阻塞通讯。
⚙️ rwkv-mobile	[Engine] The high-performance inference runtime. Supports CPU, GPU (Vulkan/Metal), and NPU. [推理引擎] 核心 C++ 运行时。支持 WebGPU, NPU, MLX 等多种硬件加速。

• Offline Inference (离线推理): Download once, run forever. No internet required.
• Multi-Modal (多模态): Chat, ASR (Speech-to-Text), TTS (Voice Cloning), and Vision (OCR/Description).
• Hardware Acceleration (硬件加速): Optimized for Qualcomm NPU, Apple MLX, and standard GPUs.
• Cross-Platform (全平台): Android, iOS, Windows, macOS, Linux.

Experience the latest RWKV-7 models on your device today. 立即在您的设备上体验最新的 RWKV-7 模型。

Special thanks to our core contributors: Molly Sophia (Core Engine), Ce Wang (App Arch), dengzi, chenqi, and the community.