hard20 pts · first answer countsmodel-selectioncost-optimizationreasoning

Reasoning model bill surprise

After switching a simple extraction task to a reasoning model, costs and latency tripled though visible outputs are the same length. Why, and what should you consider?

A.The provider made a billing errorB.Visible output got longer in tokensC.Input tokens are billed three timesD.Reasoning models generate internal thinking tokens that are usually billed and add latency; use a lower reasoning effort or a non-reasoning model for simple tasks

Discussion 0

Join free to post in the discussion.

Reasoning model bill surprise

Discussion 0

Related questions