Skip to content

Glossary entry

Frontier Model

A large AI model at the leading edge of current capability. Frontier models launch at premium prices that fall more slowly than those of older capability levels, so the economic question is when the extra capability is worth paying for.

Why it matters

Defaulting every task to the most capable model is one of the commonest causes of avoidable AI cost. Many tasks clear the quality floor on a smaller model at a fraction of the price.

Frontier models are the newest and most capable models from the leading providers. Three features shape their economics:

  • Premium pricing. The price of a fixed level of capability falls quickly over time, but each new frontier model launches at a premium, and work tends to migrate to it.
  • Reasoning. Many frontier models reason before they answer, generating reasoning tokens billed at the output rate.
  • Fast obsolescence. Today's frontier is next year's mid-tier, which matters for any commitment or tuning built around one model.

The useful question is not which model is best, but which is the cheapest setup that clears the quality floor for each task. That usually means reserving frontier models for the work that needs them and routing the rest to smaller ones, then checking that the cheaper route still meets the floor.

See The Economics of Model Selection.