Transparent pricing — proposed, not yet live.
The numbers below are launch targets we intend to validate against real measured costs. No billing is active, no payment is collected, and no one is charged today.
Proposed launch pricing — not yet live, no billing active
| Tier | Input / 1M tokens | Output / 1M tokens | Best for |
|---|---|---|---|
| Economy | $0.15 | $0.60 | Background jobs, prototypes, price-sensitive workloads |
| Standard | $0.30 | $1.20 | General chat and agent workloads |
| Priority | $0.60 | $2.40 | Latency-sensitive, interactive use |
| Batch | 50% off Standard | Large non-urgent jobs, queued and processed when capacity is cheapest | |
Prices are targets. They will only become real prices after benchmarking measures actual tokens-per-second, electricity and hardware costs on provider machines, and after the pricing engine accounts for failed requests and infrastructure expenses. If measured costs do not support a tier, the tier changes — the table does not.
Free testing during beta
During the beta, testing access is free within available resource limits. That means: use the playground and the demo API while capacity lasts, with fair-use rate limits (demo: 60 requests/minute per key). Free beta access is not a promise of unlimited free inference forever.
Provider payouts — a proposal, pending validation
The proposed split is 80% to the provider / 20% to the platform on paid completed work. This is a proposal, not a contract: it depends on measured operating costs, payment-processing fees, and whether the economics actually work at the prices above. No earnings are guaranteed. Providers are paid only for successfully completed paid workloads, and only once paid billing exists and provider compensation can actually be funded.