Distributed GPU compute for small and quick jobs with a focus on UX and cost. Run any code on any architecture, for cheap.
$
echo 'print("hi")' | curl -d @- uflop.io/run01 — How it works
Bring a job. We find a flop.
Placeholder. Submit work over a single WebSocket; the network picks the cheapest provider currently online and starts streaming output back within milliseconds. No queues. No reservations.
02 — The marketplace
Anyone with a GPU is a provider.
Placeholder. Run our agent on any machine with spare compute and it shows up as a provider. Get paid in flops, withdraw whenever.
03 — Pricing
Pay per floating-point operation.
Placeholder. Quoted per TFLOP, billed by the second. No instance types, no cold-start surcharges, no surprises on the invoice.