token.place: Decentralized Computing Power Sharing Platform – Let Idle GPUs Power AI Inference

token.place is an open-source peer-to-peer generative AI platform. It aims to connect users needing large-model inference services with volunteers contributing idle GPU computing power. It ensures communication security via end-to-end encryption, provides OpenAI-compatible APIs to lower usage barriers, builds a decentralized computing power sharing network, and promotes computing power democratization.

With the explosion of generative AI technology, the demand for large language model inference has risen exponentially. However, the centralization of high-performance GPU resources leads to high API call costs for ordinary developers, and a large amount of personal idle computing power worldwide is not effectively utilized. token.place addresses this contradiction by proposing a solution to build a decentralized computing power sharing network.

The platform consists of three roles: client (demand side), relay node (routing coordination), and computing node (computing power provider). It uses RSA+AES end-to-end encryption to ensure communication security; implements OpenAI API-compatible interfaces (v1 and v2 experimental features); supports multiple open-source models such as Llama3 and Mixtral; and has load balancing and failover capabilities.

Individuals can contribute computing power via RTX4090 gaming PCs or Raspberry Pi; enterprises can deploy private AI clusters, hybrid cloud architectures, or edge computing; it has been integrated with DSPACE's dChat feature to verify the effectiveness of the OpenAI compatibility strategy.

token.place represents a shift from centralized AI services to a distributed community-driven network. Through open-source, encryption, and compatible interfaces, it provides a path for ordinary users to participate in AI infrastructure construction. Its concept has important practical significance today when AI computing power has become a strategic resource.

Currently in the v0.1.0 development phase, the future plans include developing a Tauri desktop client to replace server.py; improving the distributed architecture; expanding multi-platform support (Windows11+NVIDIA, macOS Apple Silicon, etc.); and integrating Sugarkube to enable flexible deployment orchestration.