Hosted API for running open-source and custom ML models. Pay per second of hardware time, from $0.09/hour for small CPU to $5.49/hour for an Nvidia H100.
Developers who want to call image, video or language models through an API and pay only for compute used, or deploy a custom model without managing GPUs.
Teams that need a flat monthly price or guaranteed always-warm capacity at low cost: private models are billed for all online time including idle, and multi-GPU A100/H100 and H200 capacity requires committed spend contracts.
Pros
Cons
Most public models are billed by run time at a per-second rate set by the hardware they run on. Some models bill by input and output instead, such as per output image or per token. Private models are billed for all time the instance is online.
Listed rates include Nvidia T4 at $0.000225 per second ($0.81/hour), L40S at $0.000975 per second ($3.51/hour), A100 80GB at $0.0014 per second ($5.04/hour), and H100 at $0.001525 per second ($5.49/hour). The page shows a $ sign.
The pricing page read does not mention a free tier or free credits. Its stated principle is that you only pay for what you use. Check your account or sign-up flow for any trial terms before relying on one, as none were confirmed.
Examples on the pricing page: black-forest-labs/flux-1.1-pro at $0.04 per output image, ideogram-v3-quality at $0.09 per output image, and wan-2.1-i2v-720p at $0.25 per second of output video. Each model page shows its own estimate.
Enterprise and volume discounts are available through Replicate's enterprise program, which lists dedicated account management, priority support, higher GPU limits and performance SLAs. No enterprise prices are published, so you must contact Replicate for a quote.
Facts sourced from the vendor’s own site, last verified 2026-10-29.
The Weekly Source
One email a week. New listings, repriced tools, and the ones we would skip - including this one if its pricing changes.
