Replicate
6/10Replicate is a cloud platform that lets developers run, fine-tune and deploy open-source and custom AI models through an API, with GPUs billed by the second.
Data last updated November 2026 · Our methodology
Overview
Replicate hosts a large catalogue of community and official AI models for image, video, audio and language tasks. Cloudflare says the catalogue holds more than 50,000 models. Developers call these models through an HTTP API or the Python and JavaScript clients instead of running their own GPU servers. Teams can package their own models with the open-source Cog tool and push them as public or private models. They can fine-tune image models, and they can create deployments: private endpoints with autoscaling, configurable minimum and maximum instances, and a choice of GPU. Webhooks, streaming output and request logs are documented. There is no subscription. Most public models are billed per second of active processing, at a rate set by the hardware, from 0.09 USD per hour for a small CPU to 5.49 USD per hour for one H100. Some official models are billed per output image, per second of video or per token instead. Private models and deployments are billed for all the time an instance is online, including setup and idle time. Usage is billed monthly in arrears or drawn from prepaid credit. Enterprise contracts add volume discounts, higher GPU limits, performance SLAs, data processing agreements and a dedicated account manager. Replicate suits product developers and startups who want to add generative AI features without running GPU infrastructure. Cloudflare announced in November 2025 that it would acquire Replicate. Replicate still runs under its own brand and API, and its models are being brought into Cloudflare Workers AI. The main limitation is cold starts on shared public and community models. Reviews cite delays of tens of seconds or more on custom models. Avoiding them means paying for always-on private instances, and dedicated GPU rates are higher than on raw GPU clouds.
Our Rating
AI Software6.2/10
Hosting & Infrastructure5.9/10
Who Replicate is best for
Best software for DevelopersGood fit
Replicate is API-first, with Python and JavaScript clients, Cog packaging, webhooks and streaming, so developers can add AI models with little infrastructure work.
There is no minimum spend, billing is per second, and public models are free while idle, which suits prototyping and early products.
Product teams can try thousands of image, video and language models in the web UI before committing engineering time.
Less suited
Deployments with autoscaling suit production features, but cold starts and idle billing on private models make latency-sensitive SaaS use costly.
An enterprise tier offers SLAs, DPAs, priority support and volume discounts, but no published compliance certifications were found.
The web UI lets non-coders run image and video models, but the product is built mainly around an API and usage billing.
Pros & Cons
Pros
- Thousands of ready-to-run community and official models can be called through one API, and Cloudflare cites more than 50,000 production-ready models.
- Public models charge only for active processing time, so prototypes cost nothing while idle.
- The open-source Cog tool packages custom models for upload, and the docs cover webhooks, streaming output and fine-tuning.
- Deployments give private endpoints that autoscale from zero, with minimum and maximum instance caps to keep warm capacity or limit spend.
- Per-second hardware pricing is published openly for every GPU type, from T4 to 8x H100.
Cons
- Cold starts on shared public and community models are the most cited complaint, with reviews reporting delays of tens of seconds or more for custom models.
- Private models and deployments bill for setup and idle time, so latency-sensitive production use costs much more than the list per-second rates suggest.
- Dedicated GPU rates, such as 5.49 USD per hour for an H100, are higher than many raw GPU clouds, according to competitor comparisons.
- No SOC 2 report or public trust center was found, and enterprise SLA terms are not published.
- Bills are hard to forecast because costs vary by model, hardware and billing method, and some canceled runs are still charged.
Key Features
Not included
- Managed WordPress
- VPS Hosting
- Shared Hosting
Replicate Pricing
Enterprise
- Performance and uptime SLAs
- Volume discounts for large spend
- Higher GPU limits and reserved compute
- Data processing agreements and indemnity
- Priority support and dedicated account manager
- Multi-GPU A100, H100 and H200 capacity with committed spend contracts
Pay as you go
- Limit of 1,000 models per account
- Billed monthly in arrears or from prepaid credit
- Fast-booting fine-tunes billed only while active
- Select models can be run free for a limited time before billing setup
- Public models billed only for active processing time; setup and idle time are free
- Private models and deployments billed for all online time, including setup and idle
- No subscription and no minimum spend; billed per second of hardware time or per model output
Frequently Asked Questions
There is no fixed free tier or published free credit. Replicate lets you run select models free for a limited time, then asks you to set up billing.
Most public models are billed per second of active processing, at a rate set by the hardware, such as 0.000225 USD per second for a T4. Some official models are billed per output image, per second of video or per token. Usage is charged monthly in arrears or deducted from prepaid credit.
They run on dedicated hardware, and you pay for all the time an instance is online, including setup and idle time. Public models share a queue and charge only for active processing, but they can have cold boots.
Cloudflare announced the acquisition on 17 November 2025. Reviews from 2026 report that Replicate still runs under its own brand and API, with pricing unchanged so far, while its models are being brought into Cloudflare Workers AI.
Inputs, outputs, files and logs from predictions created through the API are deleted automatically after one hour by default. If you need to keep them, save your own copies.
Yes. You create a public or private model on the web or through the API, choose the hardware, and push it with the open-source Cog tool. Each account can have up to 1,000 models.
It adds volume discounts, higher GPU limits and reserved compute, performance SLAs, data processing agreements and indemnity, priority support, and a dedicated account manager. Prices are available from sales.
Yes. A 2026 review reports limits of 600 predictions per minute and 3,000 requests per minute on other endpoints. Replicate documents its rate limits in its predictions docs.
Similar Tools
See all alternatives
Amazon's public cloud platform, with over 200 pay-as-you-go services for compute, storage, databases, networking and AI, used by startups through large enterprises and governments.
Baseten is an AI inference platform that lets engineering teams serve, scale and train open-weight and custom models on managed multi-cloud GPUs, billed per minute.
DigitalOcean is a developer-focused cloud platform offering Droplet VMs, managed Kubernetes, databases, object storage, App Platform and GPU compute for startups and small businesses.
Ranked near Replicate in AI Software
See Replicate in the listReplicate is #240 in AI Software by our score.
