DecentralGPT operates a blockchain‑anchored peer‑to‑peer network that pools idle GPU resources globally to provide on‑demand LLM inference through a token‑priced API. The service encrypts user payloads, routes requests to the lowest‑latency node, and supports a catalog of open‑source models, enabling developers and enterprises to scale inference cost‑effectively while preserving data privacy. GPU owners receive DGC tokens proportional to their compute contribution.
Funding
Funding not disclosed
Founders
Product
Problem
Access to high‑performance AI inference is constrained by centralized cloud providers, leading to high compute costs, limited scalability, and reduced data privacy. Additionally, idle GPU resources worldwide remain underutilized, creating a supply‑demand imbalance for large language model (LLM) workloads.
Solution
DecentralGPT operates a blockchain‑anchored, peer‑to‑peer network that aggregates idle GPU capacity from participants globally to deliver LLM inference as a service. By tokenizing compute contributions, the platform aligns incentives for GPU owners while offering developers on‑demand, cost‑effective inference across a catalog of open‑source models. All user inputs are encrypted and routed through the decentralized mesh, preserving data confidentiality and eliminating single points of failure. The network’s API abstracts the underlying node topology, enabling seamless integration into existing applications without managing infrastructure.
Target Audience
The primary customers are AI developers and enterprises that require scalable, privacy‑preserving LLM inference, as well as GPU owners seeking to monetize idle compute through the network’s mining rewards.
Features
- Distributed inference layer that routes requests to the most responsive GPU node based on latency and regional proximity (e.g., USA, Singapore, Korea endpoints).
- Support for a broad model suite, including DeepSeek V3, Qwen3‑235B, Llama‑4‑Scout‑Instruct, and other open‑source LLMs, with plug‑in capability for new models.
- Token‑based pricing: $0.05 per million input tokens and $0.15 per million output tokens; optional VIP subscription at $3 / month for up to 1,000 daily conversations.
- NFT‑backed node staking that grants GPU owners proportional DGC rewards; reward allocation is calculated from real‑time compute contribution within the DBC network.
- End‑to‑end encryption of user payloads and storage on a decentralized file system, ensuring data is inaccessible to unauthorized parties.
- Transparent accounting via smart contracts, providing immutable logs of compute usage, reward distribution, and token burns for subscription fees.
- Auto‑scaling API gateway that balances load across heterogeneous hardware (NVIDIA H100, H20, etc.) while respecting minimum VRAM and bandwidth requirements.