Pricing

Replicate pricing context

Human-reviewed pricing summary paired with DevTune’s public AI search visibility benchmark.

Replicate charges on a pay-per-second model based on selected hardware tier. GPU options range from Nvidia T4 ($0.000225/sec; $0.81/hr) and L40S ($0.000975/sec; $3.51/hr) to A100 80GB ($0.001400/sec; $5.04/hr) and H100 ($0.001525/sec; $5.49/hr), up to 8× A100 configurations ($0.011200/sec; $40.32/hr) available via committed spend contracts. CPU tiers start at $0.000025/sec. Some models are billed per output unit: FLUX Schnell at $3.00/1,000 images, FLUX 1.1 Pro at $0.04/image, video models at $0.09–$0.25/second of output video, and Claude 3.7 Sonnet at $3.00/million input tokens. Private custom models on dedicated hardware are billed including idle time. Enterprise plans add dedicated account management, priority support, higher GPU limits, SLAs, and volume discounts negotiated via committed spend.