Pareto Frontier (Cost-Performance Curve)
A Pareto frontier, or cost-performance curve, plots model capability against cost to show which models give the most capability for a given price. A model sits on the frontier when no other model is at least as cheap and at least as capable, with a strict improvement on one of those measures. Builders use it to compare tiers, such as a fast cheap model against a larger flagship, and pick the trade-off that fits the task.
Also known as: pareto curve, cost-performance curve, capability-to-cost curve
On a cost-performance curve, each model is a point set by how capable it is and how much it costs to run. The models on the frontier are the best trade-offs on offer at that moment. A model is dominated when another model is at least as cheap and at least as capable, and strictly better on at least one measure. A tie on both measures does not establish dominance (McDermott, Multi-objective Optimisation, slide 29).
Labs use this framing to justify shipping several sizes of one model family, a “pro” tier and a “flash” tier, where the small model pushes the frontier toward lower cost and latency and the large one pushes it toward higher capability. For teams building on those models, the curve is an argument against defaulting to the newest or largest option.
