PROCUREMENT MECHANICS
How wholesale inference procurement works.
Specify model and volume requirements, accept a firm offer or negotiated RFQ quote, and consume reserved allowance through the unified API.
01
Define your requirements
Compare model specifications. A new purchase is at least 1 billion input-equivalent tokens. Describe your token mix, region, retention, and throughput.
02
Agree provider terms
Accept a firm offer immediately when it matches the provider’s published quantity, prices, and delivery window. Request a quote when the deal needs negotiation.
03
Consume and keep track
Use an agreement-scoped API key for supported requests. See validated usage, recorded charges, and the remaining allowance in your workspace.
Spot: capacity for now
Immediate delivery after an offer is accepted and all prerequisites hold. Your agreement still names an end time, allowance, and throughput.
Future: delivery on your schedule
Choose a start at least one hour and up to one calendar year ahead, validated at acceptance. An approved replacement can preserve unused agreed value.
YOUR NEXT BILLION TOKENS START HERE
Request capacity or pricing
Submit your model, token volume, and delivery window requirements.