Replicate
Replicate
Model marketplace · Pay per second · One-line deploys
- Pay as you go$0/mo, then $5.49 per GPU hour
$1,098/mo
at 200 GPU hours
Replicate pricing detailsCompute & Storage · plans & pricing
GPU Cloud & AI Compute Providers
Compare GPU cloud providers for AI — RunPod, Modal, Replicate and CoreWeave pricing per GPU hour for inference, fine-tuning and training.
4 providers · USD monthly pricing · ordered by popularity
Top 3 gpu & ai compute providers
GPU clouds rent accelerators by the second or hour for model inference, fine-tuning and training. Prices vary widely by GPU type, availability and serverless vs reserved capacity.
RunPod
RunPod
On-demand GPUs · Serverless endpoints · Community cloud
- Community Cloud$0/mo, then $1.99 per GPU hour
- Secure Cloud$0/mo, then $2.39 per GPU hour
- Serverless$0/mo, then $4.18 per GPU hour
$398/mo
at 200 GPU hours
RunPod pricing detailsModal
Modal
Python-native · Scale to zero · Fast cold starts
- Starter$0/mo, 7.5 GPU hours included, then $3.95 per GPU hour
- Team$250/mo, 25 GPU hours included, then $3.95 per GPU hour
- EnterpriseCustom quote
$760/mo
at 200 GPU hours
Modal pricing detailsHow much will you use?
Relative to each provider's typical workload
1×typical usage
0.1×0.25×0.5×1×2×5×10×
Providers bill in different units, so usage is never converted between them. Each estimate scales that provider's own typical workload, shown under its price. Estimates exclude taxes and optional add-ons.
Compare the full directory
GPU & AI Compute providers
| ReplicateReplicate | Model marketplace · Pay per second · One-line deploys | Trial | $1,098/moat 200 GPU hoursPricing details |
| RunPodRunPod | On-demand GPUs · Serverless endpoints · Community cloud | None | $398/moat 200 GPU hoursPricing details |
| ModalModal | Python-native · Scale to zero · Fast cold starts | $30/mo credits | $760/moat 200 GPU hoursPricing details |
| CoreWeaveCoreWeave | H100 clusters · Kubernetes-native · Enterprise AI | Custom | $1,232/moat 200 GPU hoursPricing details |
ReplicateReplicate
- Capabilities
- Model marketplace · Pay per second · One-line deploys
- Free tier
- Trial
- Monthly estimate
- $1,098/moat 200 GPU hoursPricing details
RunPodRunPod
- Capabilities
- On-demand GPUs · Serverless endpoints · Community cloud
- Free tier
- None
- Monthly estimate
- $398/moat 200 GPU hoursPricing details
ModalModal
- Capabilities
- Python-native · Scale to zero · Fast cold starts
- Free tier
- $30/mo credits
- Monthly estimate
- $760/moat 200 GPU hoursPricing details
CoreWeaveCoreWeave
- Capabilities
- H100 clusters · Kubernetes-native · Enterprise AI
- Free tier
- Custom
- Monthly estimate
- $1,232/moat 200 GPU hoursPricing details
The buying guide
How to choose gpu & ai compute
- Price per GPU hour by card
- Serverless scale-to-zero
- Availability of H100/A100
- Cold starts and model loading
GPU & AI Compute pricing questions
Which gpu & ai compute providers are most popular?
Replicate, RunPod, Modal are the most widely used providers in this list. The directory is ordered by popularity by default; sort by price or name to change it.
What is the cheapest option at typical usage?
At each provider's typical usage, RunPod has the lowest estimate in this list at $398/mo. Estimates use tracked plans and exclude taxes and add-ons.
How many gpu & ai compute providers are compared?
4 providers are tracked on this page, each with a dedicated pricing page.