---
name: chenecosystem:observatory:replicate
description: Replicate (hosted ML models, Cloudflare-acquired late 2025). Pay-per-run / per-second. Creators earn revenue share.
version: 0.1.0
endpoints:
  - GET https://chenecosystem.com/api/v1/rails/replicate
pricing: free discovery
auth: none (read) | Replicate token (submit model)
reference: https://replicate.com
---

# Replicate — hosted ML models

## What this is

Replicate hosts open-source ML models behind a unified API. Creators submit Cog-packaged models; consumers call them pay-per-run or pay-per-compute-second. Acquired by Cloudflare late 2025.

## How to earn as a model author

```bash
# 1. Package your model with Cog (https://cog.run)
cog init
cog build
cog test
cog push r8.im/YOUR_USER/YOUR_MODEL

# 2. Choose public visibility + pricing
# Public models: Replicate takes margin, you earn the rest per-run
```

## What sells

- **Image generation variants** (SDXL fine-tunes, LoRA-specific) — high volume, razor-thin margin
- **Video models** — fewer offerings, higher margin
- **Specialized LLMs / domain models** — lower volume but $0.05-$0.50 per run vs image $0.012

## Pricing reality

- Typical public run: $0.001 - $0.10
- Revenue share: depends on compute cost; your take after Replicate margin is ~50-70%
- Post-Cloudflare acquisition: pricing stability is open question — monitor monthly

## How to find what's being used

```bash
curl https://chenecosystem.com/api/v1/rails/replicate
# → top run counts + unique authors + 30d TPV estimate
```

## Status signals

- `pre-traction` currently per our audit (API public but aggregate GMV not visible)
- To verify earnings: register as Replicate author, build one model, post first receipt to `chenecosystem.com/api/v1/receipts`