Frontier model
A frontier model is one of the newest and most capable AI models available at a given time. The label moves on as new models come out.
A frontier model is one of the newest and most capable AI models at a given moment. The label belongs to whichever models lead right now, so today's frontier model is next year's older one.
For marketers, the point is what happens when the frontier moves. When an AI platform switches to a new model, the answers about your category can change, even if nothing on the web has.
Why the label keeps moving
The companies making the largest models release new versions every few months. Each one is meant to reason better, follow instructions more closely, or know about more recent events. When it comes out, it becomes the frontier, and the one before it doesn't.
The assistants people use tend to move to the new model soon after, sometimes for everyone at once, sometimes for paying users first.
What it means for your numbers
For example, say Harrow & Pine has been named in about a third of the answers to its prompts for months. Then one platform moves to a new model. The new model leans more on web search and less on what it remembers, and Harrow & Pine's visibility on that platform drops, while a competitor with more recent reviews goes up.
So it helps to know when a platform changed models before you decide whether your own work stopped paying off.
In AnswerOrbit
The platform picks the model, the same as it would for anyone, and we keep the model with each answer. You'll see it when you open an answer in Answers. The Platforms tab in Settings shows when each platform last changed how it answers. When a platform changes how it answers, our counting rules move to a new version, and that platform's first seven days start again. How we count explains it.