20
- Pay as you go
- $110
hosting & compute pricing
by Replicate
Replicate — run open models via API, billed per second of hardware.
Plans at a glance
Estimated from public pricing
Cost calculator
Pay as you go estimate
$1,098/mo
$0/mo, then $5.49 per GPU hour
Estimate covers the main usage meter only; taxes and add-ons are excluded.
| 20 | $110 |
|---|---|
| 100 | $549 |
| 200 | $1,098 |
| 1,000 | $5,490 |
| 4,000 | $21,960 |
Cheapest plan per volume highlighted. "—" means the plan does not allow that volume.
| Bandwidth / egress | Included |
|---|---|
| Builds & deploys | Cog |
| Compute | Per-second GPU |
| Regions | US |
| Storage | N/A |
| Support | |
| Team members | Orgs |
| Uptime SLA | Not included |
Replicate bills per second of GPU time for custom models, or per output for some official models.
Source: official Replicate pricing page. Estimated. Prices change — confirm before buying.