Frontier Model
A model at the current leading edge of capability.
The most capable models available at a given moment, typically the newest and largest from a handful of labs. The label is relative and moves: today’s frontier is next year’s mid-tier.
Frontier models are also the ones front-loading safety work — red teaming, evaluations, deployment policies — because their capability is where the novel risks show up first.
Treat the label as a moving reference point with real consequences attached: policy frameworks define capability thresholds that trigger heavier safeguards, so "frontier" is now partly a governance category rather than a pure performance description.
Frontier is a procurement question
Deciding whether a task needs the most capable model available.
Most production volume — classification, extraction, routing, summarisation — runs fine on a mid-tier model at a fraction of the cost. Reserve the frontier model for the minority of requests where reasoning depth changes the outcome, and route between them explicitly.
- blogIntroducing Claude 4 ↗
Hybrid models with near-instant and extended-thinking modes, reported with coding and agentic evaluations — a concrete picture of what "reasoning model" means in a product.
- reportAnnouncing our updated Responsible Scaling Policy ↗
Capability thresholds with graduated safeguards and documentation duties, written in the safety-case style: here is the evidence that this risk is adequately handled.
- reportUpdating the Frontier Safety Framework ↗
DeepMind’s front door for its frontier-risk framework of critical capability levels and mitigations. Treat as the official entry point and read the linked report for thresholds.
Foundation Model · Base model
A large model trained broadly, meant to be adapted to many downstream tasks.
Model Routing
Sending each request to the cheapest model that can handle it.
Evaluation · Evals
How you know the system is actually working.
Red Teaming
Adversarially trying to make the system do the wrong thing.