Replicate prices 3 of the models in this index. Billing model: Mixed: some models flat per-output, many billed by GPU-seconds — the same model can cost 30–50% more or less than fal.ai depending on runtime.
Replicate pricing
Hosted inference with GPU-time billing on many models, which makes per-output cost depend on generation speed.
How does Replicate bill?
Mixed: some models flat per-output, many billed by GPU-seconds — the same model can cost 30–50% more or less than fal.ai depending on runtime.
Official pricing page: https://replicate.com/pricing
Every model Replicate prices
| Model | Normalized price | Provider's own unit | vs official API | Status |
|---|---|---|---|---|
| Kling 2.5 Turbo Provideo | $0.07 /sec (720p) | $0.35 per 5s clip | — | Estimated2026-08-28 |
| Veo 3.1video | $0.40 /sec (720p) | $0.40 per second | 0% | Estimated2026-08-28 |
| Wan 2.2video | $0.021 /sec (720p) | GPU time (~$0.10 per 5s clip) | — | Estimated2026-08-28 |
"vs official API" compares this provider's normalized price to the model maker's own metered API price where both are in the index — the reseller-markup view. A negative number means this channel undercuts the official rate.