The data, as JSON
Everything on this site is available as free JSON: no API key, no rate-limit games. Use it in dashboards, routers, CI cost checks, whatever. We ask one thing: link back to llmhosting.ai where the data appears.
Regenerated with every data refresh ยท currently 2026-08-23Endpoints
| Endpoint | Contents |
|---|---|
| /api/prices.json | Per-model LLM prices across every provider serving it ($/1M tokens, cache rates, context) |
| /api/gpu-prices.json | GPU rental quotes per card and provider ($/hr, tier, live offer counts) |
| /api/gpu-history.json | Daily GPU price observations since July 2026: the growing time series behind trends |
| /api/gpu-history.csv | The same history series as CSV, for spreadsheets and pandas |
Machine-readable spec: /api/openapi.json (OpenAPI 3.1).
Example
curl -s https://llmhosting.ai/api/gpu-prices.json \
| jq '.gpus[] | select(.gpu == "H100 SXM") | .quotes[0]'Terms, in one paragraph
The normalized dataset is licensed CC BY 4.0: free for any use with attribution (a visible link to llmhosting.ai). Upstream sources keep their own licenses: LiteLLM price data is MIT, Helicone cost data is Apache 2.0, marketplace quotes reflect Vast.ai and RunPod public APIs at the stamped time. Prices are informational. Always confirm with the provider before committing spend.
Live badges for your README
Every GPU page has an embeddable SVG badge showing today's rental floor. It updates itself with each daily refresh. Grab the markdown from any GPU page, e.g. the H100 SXM:
