A single price figure combining input and output token rates at a standard 3:1 ratio — the fairest way to compare model pricing.
LLM APIs price input and output tokens separately, and output usually costs 2–5× more. Blended cost collapses the two into one number — (3 × input price + output price) ÷ 4 — modeling a typical workload where prompts are longer than responses.
It exists because comparing on input price alone is misleading: a model with cheap input and expensive output can cost more in practice than one with flatter pricing. When two models' blended costs are close, look at your actual input:output ratio — chat skews toward input, generation toward output.
modelgrep shows blended cost on the pricing table, and the calculator computes exact costs from your own token counts instead of the 3:1 assumption.