Frontier model

Definition

A term used for AI models at or near the leading edge of currently available capability.

Put simply

One of the most capable models available at that point in time.

What boards and leaders should understand

Frontier is relative. Today’s frontier model will not remain at the frontier indefinitely.

It also does not mean best for every task. Organisations increasingly need to consider capability alongside cost, latency, privacy, reliability and deployment requirements.

AI strategies based primarily on gaining access to the newest model may therefore create little durable advantage.

Questions to consider

  • If competitors can access the same frontier models, what creates durable advantage in our own AI capability?