Snapshot 2026-10-31T07:09:11+00:00. Prices are live rentable offers from the Vast.ai public feed, normalised to USD per GPU-hour. Uptime is Vast.ai's reliability score (proxy, not a measured SLA).
Cheapest now: A100 PCIE at $0.2681/GPU-hr (Florida, US); 24 GPU-hours ≈ $6.44. Get the decision report ($0.10), free preview (no wallet), or raw JSON ($0.01).
Current 24h A100 spread: $20.19 between the cheapest ($0.2681/GPU-hr) and most expensive ($1.1093/GPU-hr) live A100 offers. A wrong pick can cost more than the $0.10 report.
| Network | GPU | USD/GPU-hr | Uptime proxy | Region |
|---|---|---|---|---|
| vast.ai | A100 PCIE | $0.2681 | 0.998 | Florida, US |
| vast.ai | A100 PCIE | $0.4004 | 0.988 | California, US |
| vast.ai | A100 PCIE | $0.4007 | 0.988 | California, US |
| vast.ai | RTX 4090 | $0.4011 | 0.961 | Thailand, TH |
| vast.ai | RTX 4090 | $0.4269 | 0.980 | South Korea, KR |
| vast.ai | RTX 4090 | $0.4422 | 0.972 | Thailand, TH |
| vast.ai | A100 PCIE | $0.4685 | 0.974 | Spain, ES |
| vast.ai | A100 PCIE | $0.4704 | 0.974 | Spain, ES |
| vast.ai | RTX 4090 | $0.4711 | 0.997 | California, US |
| vast.ai | A100 PCIE | $0.5344 | 0.986 | Japan, JP |
| vast.ai | RTX 4090 | $0.8022 | 0.993 | Texas, US |
| vast.ai | A100 PCIE | $0.8672 | 0.973 | United Kingdom, GB |
| vast.ai | A100 PCIE | $1.0011 | 0.993 | United Kingdom, GB |
| vast.ai | A100 SXM4 | $1.0722 | 0.973 | Oregon, US |
| vast.ai | A100 SXM4 | $1.1093 | 0.991 | Czechia, CZ |
| vast.ai | RTX 4090 | $1.3370 | 0.999 | Texas, US |
| vast.ai | H100 SXM | $2.2022 | 0.972 | Idaho, US |
| vast.ai | H100 SXM | $2.4007 | 0.989 | Japan, JP |
| vast.ai | H100 NVL | $2.4015 | 0.999 | South Korea, KR |
| vast.ai | H100 SXM | $2.4015 | 0.989 | Japan, JP |
| vast.ai | H100 NVL | $2.5356 | 0.984 | Virginia, US |
| vast.ai | H100 NVL | $3.3049 | 0.981 | Japan, JP |
| vast.ai | H100 NVL | $3.3222 | 0.986 | Virginia, US |
Open the free structured preview (no wallet): one current cheapest offer, the live offer count, and the paid-report link. The full report adds the recommended offer, up to four alternatives, cost scenarios for 1/8/24/168/720 hours, the 7-day history, the price spread, and a rent / caution / hold decision with confidence and a basis.
Free summary: this page. Paid JSON: GET /v1/gpu/prices?model=A100 — $0.01/call. Decision report: GET /v1/gpu/report?model=A100&hours=24 — $0.10/call. Both settle in USDC on Base via x402; no account or API key. Add &format=markdown for an LLM-readable report.
Data source: Vast.ai public bundle feed. Prices change; confirm availability before renting. OpenAPI · llms.txt