Catalog
List GPU listings
Live GPU availability across every upstream provider, paginated.
GET
Filter to a specific canonical model code, e.g.
RTX_4090, H100_SXM_80GB,
B200. Combine with minVramGb to narrow further.Minimum VRAM in GB. Excludes e.g. 40GB A100 variants when set to
80.Filter to listings offering exactly this GPU count:
1, 2, 4, 8.ISO country code (
US, DE, JP, etc.) or one of us, eu, apac
to filter by broader region.Cap on hourly rate. Listings above this aren’t returned.
Pagination cursor from a previous response’s
nextCursor.How many listings per response. Max 200.
Array of listing objects, sorted cheapest-first.
Pass to the next request as
?cursor=…. null when there are no more
results.What’s a “listing”?
A single offering from one upstream provider: “this provider, in this region, has 1× H100 80GB SXM available at $2.80/hr”. Multiple listings can exist for the same model — that’s how the dashboard finds the cheapest H100 across all providers. When you callPOST /v1/rentals, the server picks the cheapest live
listing matching your spec automatically. You don’t usually need to call
/listings directly — it’s there for clients that want to display the
breakdown.
Display-price override
The dashboard’s Create page uses display-only prices (e.g. RTX 4090 is shown at 0.71/hr). This is a UI-layer override that does NOT affect the actual rental cost — when you launch, the backend bills at the real listing rate. The override exists during early launch while we hand-pick promotional pricing. API consumers get the realpricePerHourCents. Use that for
budgeting / scripts.