Frontier Model
The most capable AI models available at any given time. The models commoditize. The governance layer on top does not.
What is a frontier model?
A frontier model is the most capable AI model available at a given point in time. As of 2026, this includes Anthropic’s Claude Opus, OpenAI’s GPT-4o and successors, and Google’s Gemini. “Frontier” distinguishes these from smaller, less capable models used for specific tasks.
Why the distinction matters
Frontier models get cheaper, faster, and more interchangeable over time. What was cutting-edge six months ago becomes mid-tier. What cost $20 per million tokens drops to $3.
This commoditization is the key insight for governance: the model is rented. It’s interchangeable. The durable value layer sits on top — the corrections, the voice, the domain knowledge, the decision patterns that shape how the model behaves for a specific user.
The governance implication
A governance system built on top of frontier models must be portable. If it only works on Claude, it dies when Claude’s pricing changes or a competitor ships something better. Corrections written contrastively (incident + rule) survive model swaps. Prompt exploits tied to specific model behavior don’t.
The model is the engine. The kernel is the driver. The engine gets replaced every few years. The driver’s experience compounds.