Active development. APIs and on-chain layouts may change. Track on GitHub.
S SolanaLM
Home / Use cases / DePIN GPU compute that earns

DePIN GPU compute that earns

Put idle DePIN or spot GPUs to work as an OpenAI-compatible inference node — dual SOL revenue from inference requests and federated training rounds.

The problem

DePIN networks and GPU marketplaces make cheap idle hardware abundant, but they rent it by the hour and leave the serving stack, routing, and billing to you. Generic compute is not a monetized, OpenAI-compatible endpoint — and light or bursty traffic still pays for reserved hours.

How SolanaLM does it
  • A node is a container that registers on its first heartbeat, so it drops straight onto DePIN networks (io.net, Akash, Render, Nosana) or marketplace/spot GPUs (Vast.ai, RunPod) — wherever inference-capable hardware is cheapest.
  • Nodes earn SOL per served inference request and per federated training round — two revenue streams from the same hardware.
  • Per-request micropayments with ~0.4s Solana finality and ~$0.00025 fees mean you get paid as work completes, not on a monthly invoice — and callers can be autonomous agents paying per call.
  • A single 24GB consumer/prosumer card serving a quantized 7B–8B model via the llama.cpp GGUF backend is enough to start: a few dollars a day of hardware against per-request revenue.

Try it for depin gpu compute that earns.

Read the deep dive, then run the quickstart.

Other use cases