token pricing
The cost model where LLM APIs charge per token consumed (input and output), not per request or per hour.
Think of it like
Cell phone billing—pay for minutes (tokens) used, not a fixed monthly fee.
Example
Claude API: $3 per 1M input tokens, $15 per 1M output tokens. A 1,000-token response costs $0.015.
How it actually works
Token pricing aligns incentives—longer responses cost more, so you think about efficiency. But it's opaque (users don't know how many tokens their text will tokenize to). Output tokens usually cost more than input (generation is harder). Pricing pressure from competition.
For product teams
Transparent and usage-based; scales naturally with demand.
For engineers
Tokenization adds latency; cache/batch pricing is becoming differentiator.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome