GPU cloud · 4 GPU models tracked
Replicate
$0.81–$5.49per GPU-hour, the cheapest row to the dearest
Models run on hardware billed by the second - public models for the seconds a request runs, private ones for all the time their instances are up - shown here as an hour; each price is an instance with its CPU and memory. H200 and 4- or 8-GPU machines are sold by contract and are not listed.
As of , Replicate lists 4 GPU models we track, from $0.81 per GPU-hour (NVIDIA T4, on-demand) to $5.49 (NVIDIA H100 80GB (form factor not stated), on-demand).
| Website | replicate.com/ |
|---|---|
| Pricing page | replicate.com/pricing |
| Regions and tiers | No region or tier named in the rates below: one global rate card |
| Models tracked | 4 |
Price by GPU
As of , per GPU-hour: Replicate’s on-demand rate for each card it lists, a cheaper market it also sells, and where the on-demand rate stands among the providers we track.
| GPU | Replicate on-demand, per GPU-hour | Against the market (median rate card) |
|---|---|---|
| NVIDIA H100 80GB (form factor not stated) | $5.49 captured Deploy ↗ | 46% above the median $3.75 · 28 providers 23 of 28 list lower · lowest Gcore $1.72 · compare |
| NVIDIA A100 80GB | $5.04 captured Deploy ↗ | 153% above the median $1.99 · 63 providers 60 of 63 list lower · lowest Yotta Labs $0.92 · compare |
| NVIDIA L40S | $3.51 captured Deploy ↗ | 147% above the median $1.42 · 40 providers 38 of 40 list lower · lowest CloudRift $0.63 · compare |
| NVIDIA T4 | $0.81 captured Deploy ↗ | 47% above the median $0.55 · 19 providers 16 of 19 list lower · lowest immers.cloud $0.24 · compare |
The same, in words
- Replicate lists the NVIDIA H100 80GB (form factor not stated) from $5.49 per GPU-hour on-demand. The on-demand rate is unchanged since first read, Sep 23, 2026 (7-day change: none). Against the market: 46% above the median rate card ($3.75 across 28 providers); 23 of the 28 providers we track list a lower on-demand rate, the lowest Gcore at $1.72 (Gcore’s rates, Replicate vs Gcore).
- Replicate lists the NVIDIA A100 80GB from $5.04 per GPU-hour on-demand. The on-demand rate is unchanged since first read, Sep 23, 2026 (7-day change: none). Against the market: 153% above the median rate card ($1.99 across 63 providers); 60 of the 63 providers we track list a lower on-demand rate, the lowest Yotta Labs at $0.92 (Yotta Labs’s rates, Replicate vs Yotta Labs).
- Replicate lists the NVIDIA L40S from $3.51 per GPU-hour on-demand. The on-demand rate is unchanged since first read, Sep 23, 2026 (7-day change: none). Against the market: 147% above the median rate card ($1.42 across 40 providers); 38 of the 40 providers we track list a lower on-demand rate, the lowest CloudRift at $0.63 (CloudRift’s rates, Replicate vs CloudRift).
- Replicate lists the NVIDIA T4 from $0.81 per GPU-hour on-demand. The on-demand rate is unchanged since first read, Sep 23, 2026 (7-day change: none). Against the market: 47% above the median rate card ($0.55 across 19 providers); 16 of the 19 providers we track list a lower on-demand rate, the lowest immers.cloud at $0.24 (immers.cloud’s rates, Replicate vs immers.cloud).
All listed prices
One table per price type, never mixed. Every rate normalized to one GPU for one hour; a price opens the page it was read from, and 7d marks a rate that moved in the past seven days. The day each figure was first listed is in this page’s data, below.
On-demand
| GPU · what is sold | $ / GPU-hour | Read | Source |
|---|---|---|---|
| NVIDIA T4 | $0.81 | captured | Deploy ↗ |
| NVIDIA L40S | $3.51 | captured | Deploy ↗ |
| NVIDIA A100 80GB | $5.04 | captured | Deploy ↗ |
| NVIDIA H100 80GB (form factor not stated) | $5.49 | captured | Deploy ↗ |
Ask an AI agent about Replicate’s rates: Claude, Cursor, VS Code and ChatGPT can read them through the GPU Economy MCP server - free, no key, every answer citing the page and read time.
Every rate on this page as data: /providers/replicate/prices.json — the same fields as prices.json, narrowed to this provider. CC BY 4.0, no key, no rate limit. On your own page: embed this table (/embed/providers/replicate, an iframe, no script).
Compare Replicate with
Every provider that lists at least two of the same GPUs on-demand, most shared GPUs first; each opens the side-by-side page.
- Cerebrium 4 shared GPUs
- E2E Networks 4 shared GPUs
- EmpirioLabs 4 shared GPUs
- Hugging Face 4 shared GPUs
- Lightning AI 4 shared GPUs
- Amazon Web Services 3 shared GPUs
- Bamboozle Web Services 3 shared GPUs
- Baseten 3 shared GPUs
- Gcore 3 shared GPUs
- Hikube 3 shared GPUs
- immers.cloud 3 shared GPUs
- Koyeb 3 shared GPUs
- Lyceum 3 shared GPUs
- Massed Compute 3 shared GPUs
- Modal 3 shared GPUs
- Sesterce 3 shared GPUs
- airgpu 2 shared GPUs
- Beam 2 shared GPUs
- Civo 2 shared GPUs
- Cloud.ru Evolution 2 shared GPUs
- CloudRift 2 shared GPUs
- CoreWeave 2 shared GPUs
- Crusoe 2 shared GPUs
- Cudo Compute 2 shared GPUs
- Deep Infra 2 shared GPUs
- Geodd 2 shared GPUs
- Google Cloud 2 shared GPUs
- GPU.ai 2 shared GPUs
- HexGrid 2 shared GPUs
- Hostinger 2 shared GPUs
- Inferless 2 shared GPUs
- Infomaniak 2 shared GPUs
- Krutrim Cloud 2 shared GPUs
- Leafcloud 2 shared GPUs
- Lium 2 shared GPUs
- Microsoft Azure 2 shared GPUs
- MimicPC 2 shared GPUs
- Northflank 2 shared GPUs
- Oracle Cloud (OCI) 2 shared GPUs
- OVHcloud 2 shared GPUs
- packet.ai 2 shared GPUs
- QuantaCloud 2 shared GPUs
- RunComfy 2 shared GPUs
- RunPod 2 shared GPUs
- Runware 2 shared GPUs
- Seeweb 2 shared GPUs
- STACKIT 2 shared GPUs
- Vast.ai 2 shared GPUs
- Verda 2 shared GPUs
- VESSL AI 2 shared GPUs
- Yandex Cloud 2 shared GPUs
- Yotta Labs 2 shared GPUs
Data updated: