{
  "slug": "replicate",
  "name": "Replicate",
  "url": "https://replicate.com",
  "cat": "ai",
  "group": "ai",
  "tagline": "Per-run model hosting API",
  "price": "Per second of compute (T4 $0.000225/s, A100 80GB $0.0014/s) or per output for media models (FLUX Dev $0.025/image, video from $0.09/s). Some LLMs bill per token. No free plan.",
  "tier": "paid",
  "pick": false,
  "what": "A catalog of thousands of community and official models — image, video, audio, LLM — each behind one prediction API. It is an <b>inference host</b>: you push a Cog container or run someone else's, and billing is per second of GPU time or per generated output.",
  "why": "The fastest way to call a weird open model — a niche upscaler, a voice clone, a brand-new video model — without touching CUDA. Public models cost nothing when idle, so hobby projects sit at $0.",
  "warn": "Private and dedicated models bill for boot and idle time, not just inference — a deployment left running overnight can cost more than a month of API calls. Cold starts on rare models take minutes.",
  "install": "pip install replicate (or npm i replicate)",
  "category": {
    "key": "ai",
    "name": "Model APIs, routers & observability",
    "desc": "Providers, gateways, inference hosts and eval layers — constantly conflated, listed separately here.",
    "group": "ai",
    "groupName": "Put AI inside it"
  },
  "freshness": null,
  "health": {
    "ok": true,
    "status_code": 200,
    "final_url": "https://replicate.com/",
    "checked_at": "2026-08-18T23:43:12.623+00:00"
  },
  "corrected_at": null,
  "upstream_changed_at": null
}