Three things make an Aigarth ANN different from any model you can download. It is owned, it is billable, and it is discoverable by the network.
An ANN is a versioned neural network: weights, training data, benchmarks: all signed and published on Qubic.
License terms are enforced by the chain. Every call pays you, your co-creators, and your stakers automatically.
Set your price. Stake QUBIC to be discoverable. The network routes demand to the most accurate, most staked ANNs.
No third-party tooling. No rented GPUs. The studio, the marketplace, the network: one stack, one chain, one revenue line.
Start from any base model. Vision, text, audio, embeddings, agents. Run distributed training on the same network that serves inference.
Every revision is a separate ANN with its own benchmarks. Roll back, branch, compare. The lineage is signed and public.
List on the marketplace. Set a price, set a stake requirement, set a license. The studio handles signing, benchmarks, revenue splits.
Open, commercial, or restricted. Per-region, per-tenant, per-call. Composable licenses enforced by the network, not a PDF.
OpenAI-compatible API. Sub-50ms latency across 47 regions. The same network serves your ANN and routes traffic to the best nodes.
75% to creator, 15% to stakers, 10% protocol. Revenue is distributed in real time. No invoices, no intermediaries.
Every creator on Aigarth follows the same loop. No exceptions, no shortcuts. Stake, train, publish, earn, expand.
Qubic wallet only. No KYC, no email. Five seconds.
Your stake = your compute budget. Longer stake, lower fee.
Pick a base. Add your data. Train on the network. Publish when ready.
Set price, license, stake requirement. Live in 60 seconds.
Every call pays you + your stakers. Idle capacity earns too.
Your stake reserves compute. Your compute trains an ANN. Your ANN earns on every call. Idle capacity earns on the open market. The cycle never stops.
1,247 ANNs live. These are the ones earning the most this week. Browse, stake, license, or fork.