Token-based pricing
Token-based pricing is a billing model that charges customers per unit of text (token) processed by an AI model, with separate rates for input and output tokens, mirroring how most LLM vendors themselves price API access.
How token-based pricing is structured
A token-based price list specifies two rates — one for input (prompt) tokens and one for output (completion) tokens — usually quoted per 1,000 or per 1,000,000 tokens. A company selling an AI feature can pass vendor token pricing straight through, apply a markup on top of vendor cost, or blend token pricing into a different customer-facing unit (per message, per document, per seat) while still metering and reconciling cost at the token level internally.
Why it is the dominant model for AI features
Token-based pricing keeps customer price closely aligned with the underlying cost driver, since nearly every upstream model vendor itself bills in tokens. That alignment is what makes it possible to protect gross margin as usage scales — a customer-facing price with a fixed markup over token cost holds margin consistently, whereas a flat per-seat price with no usage ceiling can be undercut entirely by a handful of heavy users.
Worked example
A customer is billed $4 per 1,000,000 input tokens and $12 per 1,000,000 output tokens. A single chat session uses 5,000 input tokens and 1,200 output tokens.
Customer charge for one session
- Input charge = 5,000 tokens × ($4 ÷ 1,000,000) = $0.0200
- Output charge = 1,200 tokens × ($12 ÷ 1,000,000) = $0.0144
- Total customer charge = $0.0200 + $0.0144 = $0.0344