Any model, served by the network.

Chat, embeddings, images, audio and video through the API shape your code already speaks. Point your client at the network and keep your code.

Prefix and state reuse for dense, sliding-window, hybrid recurrent and state-space, mixture-of-experts, multimodal and media models, not a special case for one model.

Many operators serve the same model. Filter providers by price, latency, location and the trust credentials you accept.

Pay per request or per token in TNZO or stablecoins, settled on chain. No minimum commitments.

Providers publish signed policies. Weights are verified by hash, and confidential workloads can run in attested hardware.

Large models split across machines and clusters, so models bigger than one GPU can still be served by the network.