Federated model training
Coordinate training across nodes that never share raw data, with FedAvg, FedProx, FedAdam, and SCAFFOLD.
Training on data that cannot leave its owner — hospitals, banks, edge devices — rules out centralizing a dataset. Rolling your own federated coordinator means handling non-IID data, dropped clients, and privacy accounting yourself.
- A federation runtime ships FedAvg, FedProx, FedAdam, and SCAFFOLD, so you pick the aggregation strategy that fits your data skew.
- Non-IID data handling, circuit breakers, and automatic recovery on dropped clients keep rounds progressing.
- Secure aggregation and (epsilon, delta)-differential privacy are on by default, with configurable noise for stricter regimes.
- Nodes that only train — no inference — still earn SOL from round participation.
Try it for federated model training.
Read the deep dive, then run the quickstart.
Other use cases
Privacy-preserving inference
Route confidential prompts through onion circuits so no single node sees both source and content.
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.
Machine-payable inference for agents
Give an autonomous agent an OpenAI-compatible endpoint it can pay for per call in SOL — point the existing SDK at the gateway, keep your app code unchanged.