A designer tool generating images
- Create a Replicate API token
- Call black-forest-labs/flux-1.1-pro
- Pay $0.04 per output image
needs API key
Replicate (a Cloudflare company) · Model hosting and inference API
Replicate
Run and deploy open-source and custom AI models through an API, paying per second of hardware or per output.
- Provider
- Replicate (a Cloudflare company)
- Latest confirmed model
- Not applicable / not verified
- Pricing
- Paid only
- Free plan
- No
- API
- Yes
- Status
- Generally available
- Last verified
- Oct 19, 2026
What is Replicate?
Replicate hosts thousands of public AI models for images, video, audio and text that you can call with a simple API, and lets you deploy your own models. Most public models are billed only for processing time on the chosen GPU, while some official models are billed per token or per output image. Cloudflare acquired Replicate, which continues as a separate brand.
Using or deploying AI models, especially image and video models, without managing GPUs.
Why use Replicate?
Practical benefits
- Large catalogue of ready-to-run models
- Per-second billing on public models
- Deploy custom or fine-tuned models
- Simple API
Who it suits
- Developers adding image or video generation
- Startups testing many models
- ML engineers deploying custom models
When another tool may fit better
- Users needing a documented free tier
- Very high steady workloads where reserved GPUs may be cheaper
- Non-technical users
- Learning curve
- Some experience helps
- Skills needed
- Basic API usage; ML knowledge for custom deployments.
Example workflows
- Pick a model page, copy the API snippet, run it with your API token, pay per run
What can it do?
- Public model catalogueRun community and official models via API.
- Per-second hardware billingCPU from $0.09/hr to H100 at $5.49/hr.
- Per-output pricingSome official models priced per image or per token.
- Private deploymentsDedicated hardware billed for setup, idle and active time; fast-booting fine-tunes bill only active time.
Good for
- Image generation in apps
- Video and audio generation
- Hosting fine-tuned models
- Prototyping with many models
Where it runs
Pricing and plans
Paid only Usage-based. Public models mostly bill by hardware time (e.g. Nvidia T4 $0.81/hr, A100 80GB $5.04/hr, H100 $5.49/hr, billed per second). Some official models bill per output or token, such as flux-1.1-pro at $0.04 per image.
| Plan | Price | Billing | Free trial / tier | Usage limits | Main inclusions | Model access | API access | Restrictions |
|---|---|---|---|---|---|---|---|---|
| Pay as you go (public models) | Per second of hardware time, or per output/token for some official models | usage-based | — | — |
|
Public models | Yes | — |
| Private deployments | Per second of hardware time including setup and idle (fast-booting fine-tunes bill active time only) | usage-based | — | — |
|
Your models | Yes | — |
API and developer access
- Official API
- Yes
- Documentation
- API docs (opens in a new tab)
- Authentication
- API token (Bearer)
- Pricing model
- per second of hardware; per output or per token for some models
- Free API quota
- None documented — the free plan above (if any) is for the app, not the API
- Rate limits
- Not documented
- Context / input limits
- Not documented
- SDKs
- Not documented
- Streaming
- Not documented
- Batch processing
- Not documented
- Structured output
- Not documented
- Tool / function calling
- Not documented
- Commercial use
- See the provider terms
- Data retention & privacy
- See the provider privacy terms
Published API rates
| Item | Price | Unit |
|---|---|---|
| CPU (Small) | $0.09 | per hour (billed per second) |
| Nvidia T4 | $0.81 | per hour (billed per second) |
| Nvidia L40S | $3.51 | per hour (billed per second) |
| Nvidia A100 (80GB) | $5.04 | per hour (billed per second) |
| Nvidia H100 | $5.49 | per hour (billed per second) |
| black-forest-labs/flux-1.1-pro | $0.04 | per output image |
| deepseek-ai/deepseek-r1 input | $3.75 | per 1M input tokens |
A free Replicate account is not the same thing as free API access. Unless a free API quota is listed above, API calls are billed separately.
Advantages and disadvantages
Advantages
- Simple pay-per-use pricing
- Huge model selection
- Good for media models
Disadvantages
- No free tier mentioned on the pricing page
- Private deployments bill idle time
- Costs vary by model runtime and can be hard to predict
Privacy: Not verified in this research.
Real-world examples
A developer testing a community model
- Find the model on replicate.com
- Run it via API
- Pay only for GPU seconds used
needs API key
An ML team deploying a fine-tune
- Push a custom model
- Create a private deployment on L40S ($3.51/hr)
- Scale based on traffic
needs paid plan
Replicate alternatives
Hugging Face
Hub for sharing open models, datasets and demo apps, with paid compute for inference and hosting.
Replicate vs Hugging FaceTogether AI
Cloud platform for running, fine-tuning and training open models on serverless APIs or rented GPUs.
Replicate vs Together AIOpenRouter
One OpenAI-compatible API key for hundreds of models from many providers.
Replicate vs OpenRouterGroqCloud
Low-latency inference API for open models running on Groq's LPU hardware.
Replicate vs GroqCloudFrequently asked questions
Is Replicate free?
The official pricing page does not mention a free tier. You pay per second of hardware or per output depending on the model.
How is Replicate billed?
Most public models bill for processing time only, at per-second hardware rates such as $0.81/hr for a T4 or $5.49/hr for an H100.
Do private deployments cost more?
They bill for setup, idle and active time on dedicated hardware, except fast-booting fine-tunes, which bill only active time.
Who owns Replicate?
Cloudflare announced its acquisition of Replicate in November 2025; its SEC filing dates the purchase to December 1, 2025. Replicate continues as a distinct brand.
How much does an image cost?
It depends on the model; for example black-forest-labs/flux-1.1-pro is $0.04 per output image.