Regions and pricing
Live
Open in ChatGPT
(opens in new tab)Last verified: 2026-08-11Region discovery
GET /regions returns the default routing mode and each region's supported and currently available API types. Use region in the JSON body or x-modelrush-region for chat when a specific execution path is required.Region fields describe different evidence.
released_regions is Provider research, configured_regions has ModelRush credentials and endpoint configuration, healthy_regions passed a recent real check, available_regions is eligible for the current API key and runtime, and routed_region is where one request actually executed. Never merge them into a generic “supported” claim.Automatic routing
Omit the region to use
auto. ModelRush considers only healthy configured Deployments that have a valid price for their exact Region. A retryable chat failure may move to another eligible Region before a response starts; every attempt is recorded, and an explicit Region never crosses its data-residency boundary. Record the returned modelrush_region or x-modelrush-region value so the actual execution path is auditable.Pricing source
Token and non-token prices are versioned by model Operation and Region. Auto routing reserves enough credit for the highest enabled customer price among eligible retry Regions, then settles the price of the Region that actually completed the request. Evaluation prices remain visible while
billingEnabled=false, but billed usage stays $0. Use the public pricing page and model catalog as the current commercial display. Do not copy prices into application code; associate cost with the returned model, Region, and usage data.PreviousAudio APIs