Replicate Model hosting
Hardware runtime or model input and output.
Infrastructure. Metered pricing.
What the company publishes
- Pricing mechanism
Metered
- Chargeable unit
Hardware runtime or model input and output
- Included allowance
Not publicly specified
- Packaging
Usage based, with model-specific prices
- Credits or prepayment
Not publicly specified
- Overage treatment
Not publicly specified
- Enterprise exceptions
Not publicly specified
Checked Nov 3.
Source notes
The page says most models are billed by runtime at hardware-specific per-second prices.
It says some models are billed by input and output instead.
Lago analysis
- Tradeoff
Per-run charges align to actual compute, but different models make cross-model budgets less simple.
- Billing and revenue work
Meter runtime and model output with a model identifier, then rate each according to its published rule.
This is an implementation inference, not a claim about the company's internal systems or Lago usage.