Decentralized GPU compute, settled on Solana.
One order book over Nosana and Akash, rebuilt from their public feeds every 150 seconds. Spec a job, take the cheaper network, pay in USDC. On every GPU both networks quote, the Solana side is the cheaper one.
sources: nosana (— markets) · akash (— models)
The same GPU, on both networks, at the same moment.
These are the GPU classes Nosana and Akash both quote right now. Each pair is one card: the Solana price, the Cosmos price, and the gap between them. Nothing here is modelled — it is two public price feeds put on one axis.
Price a job against the live book.
Pick the hardware and the term. The quote is the cheaper network's current rate multiplied by your hours — no spread added, no term discount invented.
Recorded with the job for routing. Neither feed publishes regional rates, so it is not priced in.
Every market, unaggregated.
One row per live market. Multi-GPU nodes are normalised to a per-GPU rate so an 8×H100 box sits on the same axis as a single card.
| market | available | |||
|---|---|---|---|---|
Quoted in USDC. Settled on Solana.
Pricing is only half of it. The rail a job settles on decides how fast it clears and what it costs to clear — which is the reason the cheaper network here is also the Solana one. This is that chain, right now.