Frontier Model
Also called Frontier AI, Frontier AI Model
One of the most capable AI models available at a given time, at the leading edge of what current systems can do. The label changes as the leading edge moves.
What is a frontier model?
A frontier model is a general-purpose AI model at or near the cutting edge of capability. There is no single official definition. The term is used by labs, researchers and policymakers to point at the newest and most powerful systems, which often bring new abilities and new risks.
How does it work?
Frontier models are usually foundation models trained at very large scale, often large language models that also handle images, audio or tool use. Because the frontier moves, a model that was frontier last year may be ordinary now. Safety testing and reporting rules are often aimed at this group because their abilities are least understood.
What is a simple example?
When a lab releases a new flagship model that beats every earlier system on reasoning and coding tests, that model is described as frontier until a stronger one replaces it.
Why does it matter?
The label marks where capability, cost and risk are changing fastest. It is distinct from AGI and superintelligence, which are hypothetical: frontier models exist today but are still narrow.