Replicate Pricing Calculator 2026
Estimate your total cost including hidden fees
The Replicate calculator estimates the cost of your selected plan. Change the team size where pricing is per person, choose an available billing period, and add reported extras that apply to your contract.
Enter your requirements below to see the recurring cost and first-year total. Unverified extras are excluded.
- Plans: 3 tiers; Free–$43.92/per prediction
- Serverless Pricing Premium: $1/minute
- GPU Rental Markup: 200-300% markup over alternatives
- Managed Service Premium Over Raw GPU Compute: $3-$4/hr
Reported extras may depend on your usage and contract. Confirm the final quote with Replicate.
Replicate pricing ranges from $0 to $43.92 per per prediction as of September 2026. Replicate offers 3 pricing tiers. Reported extras are shown below. Only costs with a clear amount and billing basis can be added to the estimate. Pricing verified from 3 sources by CostBench.
Estimate Your Monthly Cost
Enter your expected monthly usage:
- Public models billed per prediction (token, image, or second)
- Private/custom models billed per second of dedicated hardware time
- A100 GPU: $0.00140/sec ($5.04/hr); H100: $0.001525/sec ($5.49/hr)
Real-World Replicate Cost Examples
Image Generation Finetuning Cost Comparison
$For a task that requires 1 H100, Replicate charges $1/minute ($60/hr), while 8xH100s on Runpod cost just $2.88/hr - making Replicate 20x more expensiveFor a task that requires 1 H100, Replicate charges $1/minute ($60/hr), while 8xH100s on Runpod cost just $2.88/hr - making Replicate 20x more expensive
Running image generator finetuning on Replicate's serverless versus alternatives shows the cost difference. Replicate charges $1/minute for workloads that could run on a single H100.
HN discussion on finetuning costsAudio Transcription at Scale (400 hours via Whisper)
$~$70 for 400 hours of audio (~$0.0029/run per 1-minute chunk)~$70 for 400 hours of audio (~$0.0029/run per 1-minute chunk)
Transcribing 400 hours of audio using Whisper Large v2 via Replicate's Pay-as-you-go inference API, processing approximately 1-minute audio chunks as individual runs.
Reddit/LocalLLaMA (2025-02-14)A100 GPU Compute Per Hour
$~$5/hr per A100~$5/hr per A100
Running a single A100 80GB GPU instance on Replicate for model inference or fine-tuning via the Pay-as-you-go plan.
HN community (2024-10-04, 2026-03-27)Compare at This Team Size
Frequently Asked Questions
01 How accurate is this Replicate pricing calculator?
This calculator uses official Replicate pricing data verified as of 2026-07-30. Hidden cost estimates are based on 4 verified cost categories from real user reports. Actual costs may vary based on negotiated discounts, specific feature requirements, and implementation complexity.
02 What hidden costs should I include in my Replicate budget?
Our calculator includes 0 verified hidden cost categories for Replicate: . Toggle each to see how they affect your total cost.
03 Should I choose monthly or annual billing for Replicate?
Compare the published annual total with twelve monthly payments when both are available. An annual commitment may reduce flexibility; confirm cancellation terms and whether the quoted plan fits your needs.
04 How do I know which Replicate tier I need?
Start with your must-have features. Replicate offers 3 tiers ranging from $0 to $43.92/per prediction. Entry tiers work for basic needs, while enterprise tiers add advanced security, customization, and support.
05 Can I negotiate Replicate pricing below calculator estimates?
Negotiated pricing depends on the vendor, plan and contract. We do not assume a discount in this estimate. See our negotiation guide for tactics.