Catalog/ № 052
Replicate
shipRun any community model via API — pay per second, no GPU wrangling.
replicate.comVerified 2026-08-15
ledger entry · observed by @whysanesanders
The fastest way to ship a model you do not want to host: thousands of community models behind one API with per-second billing. Perfect for prototyping and spiky workloads.
Known limitations
- Per-second pricing gets expensive at steady volume.
- Cold starts on less-popular models.
- Model quality and maintenance vary by community author.
- Acquired by Cloudflare (2025-11); API and pricing unchanged so far.
Facts
- pricing
- paid1
- price note
- pay per run; image models from ~$0.003/image
- free tier
- no
- open source
- no
- api
- yes
- self-host
- no
- category
- dev-infra
Sources
If we can't show the source, we don't print the number.
Same shelf — dev-infra
Fireworks AI
dev-infra
Fast inference platform for open models — serverless and on-demand GPUs.
APIFREE TIER
freemium$1 starter credits; serverless from $0.10/1M tokens; on-demand GPU from $7/hr
Groq
dev-infra
Ultra-fast inference on custom LPU hardware — open models at 500+ tok/s.
APIFREE TIER
freemiumfree tier; from $0.05/1M tokens (small models)
Hugging Face
dev-infra
The GitHub of AI — models, datasets, Spaces and inference endpoints.
APIFREE TIER
freemiumfree; PRO $9/mo; inference pay-as-you-go