# Replicate

> Replicate provides a developer platform and API to run, fine-tune, deploy, and scale machine learning models with pay‑for‑what‑you‑use hardware billing.

Replicate is a developer platform and API for running, fine‑tuning, deploying, and scaling machine learning models. It exposes models as production-ready APIs and supports running community and private models with per-second hardware billing. Teams can deploy custom models (via Cog), fine-tune models with their data, and monitor predictions with logs and metrics.

- **One-line API access** *call any model with a single API request using official SDKs (Node, Python, HTTP).* 
- **Pay‑for‑what‑you‑use billing** *models are billed by runtime (per-second) and by hardware type so you only pay for compute used.*
- **Deploy custom models** *package and deploy your own model with Cog to create a scalable API endpoint.*
- **Fine-tuning support** *train or fine-tune models on Replicate to produce custom versions for specific tasks.*
- **Hardware choices & scaling** *choose CPU or GPU hardware (T4, L40S, A100, etc.) and scale automatically when demand increases.*
- **Logging & monitoring** *built-in metrics and logs let teams track model performance and debug predictions.*

To get started, sign up on the web, obtain an API token, and use the Node/Python/HTTP SDKs to run a published model or deploy your own model packaged with Cog.

## Features
- Run community and private models via API with one line of code
- Fine-tune and train models with your data
- Deploy custom models using Cog
- Per-second, hardware-based billing (CPU and multiple GPU types)
- Logging, metrics, and automatic scaling for deployed models

## Integrations
Node.js SDK, Python SDK, HTTP API, Cog (model packaging), GitHub

## Platforms
WEB, API, DEVELOPER_SDK

## Pricing
Paid

## Links
- Website: https://www.replicate.com/
- Documentation: https://replicate.com/docs
- Repository: https://github.com/replicate
- EveryDev.ai: https://www.everydev.ai/tools/replicate