by Replicate, Inc. · founded 2019 · updated Sep 2026
Replicate turns thousands of open models, image, video, audio, LLMs, into pay-per-use APIs. Cog packaging lets anyone publish a model; per-second billing and zero infra make it the fastest path from model to product feature.
Fixed weights, sources attached, the formula
Each score is built from the claims below. This is a snapshot, so check the current terms with the vendor before you commit to anything.
1 of 3 claims below are past their re-check window. Pricing, availability and security details move fast, so confirm those with the vendor.
| Claim | Source | Checked |
|---|---|---|
| Thousands of public models runnable via a single API· re-check | Replicate model catalog | Aug 2025 |
| Cloudflare's acquisition of Replicate closed Dec 1, 2025; the API and existing models kept working unchanged, and no pricing update accompanied the close — H100 still billed at $0.001525/sec ($5.49/hr) | Replicate/Cloudflare blog; independent review coverage ↗ | Sep 2026 |
| Replicate requests can now route through Cloudflare AI Gateway at no extra cost, billed at Replicate's standard rates, as the brand continues operating independently ahead of deeper Workers AI integration | Cloudflare Community changelog ↗ | Sep 2026 |
Following its acquisition, Replicate's catalog is being integrated with Cloudflare Workers AI. Expect lower latency and new pricing tiers; watch for migration guidance if you depend on current endpoints.
Hugging Face
The GitHub of machine learning: models, datasets, Spaces and inference.
Groq
Ultra-fast LLM inference on custom LPU hardware.
Ollama, Inc.
Run open LLMs locally with one command.