Capacity loop
The compute market is crowded. The loop is not.
Flatkey connects metered model access and dedicated inference capacity, so teams can move heavy traffic without creating another disconnected vendor workflow.
Flatkey Compute
Dedicated inference GPUs, by the hour.
Move from shared per-token routing to dedicated capacity when workloads need predictable throughput, enterprise review and unified billing.
$1.79/hr
H100 market-floor target
SLA
Signed operating terms
1 bill
Models and compute together
Upgrade path
From per-token to dedicated, in one step
Start with the router, identify stable high-volume workloads, then reserve hourly GPU capacity under the same commercial relationship.
Positioning
Not another GPU cloud
Compute is designed as a B2B extension to Flatkey usage, not a separate infrastructure marketplace with another account system.