SN20 runs a rolling program — up to two new teams enter each month, and each spends six months in incubation. These are the teams running right now. Browse them, follow their progress, and back the ones you believe in.
A routing layer that directs inference requests to the best-priced, lowest-latency model across the network.
Cryptographic proofs that a model actually ran the computation it claims — trust for off-chain inference.
Turns raw farm sensor feeds into yield predictions and irrigation guidance through a network of specialist models.
An on-chain market where labelers are scored and paid for high-quality annotations on training datasets.
Streaming transaction-risk models that flag fraud in milliseconds across payment and DeFi rails.
Translation models for languages the big providers ignore — built by native-speaker miners.
Generates and validates synthetic datasets on demand — clean training data without the scraping.
A forecasting subnet for demand, energy and market signals — competitive miners ranked on accuracy.