← Learn

Frontier model

A top-tier, usually proprietary model at the current capability frontier — the reference point open-weight and smaller models are measured against.

Learn

When to use it

Use a frontier model when you need the highest performance available for a task, beyond what open-weight or smaller models can achieve. Frontier models provide a benchmark for evaluating other models and are essential for applications requiring cutting-edge capabilities, such as advanced natural language understanding or complex image recognition.

Quick example

In a recent catalog review task, a $500 fine-tuning of a 9B open model outperformed a frontier AI model, showcasing the potential of task-specific optimization. The frontier model served as the benchmark, demonstrating its role as the standard for measuring improvements. This scenario illustrates how frontier models act as a reference point for evaluating the effectiveness of customized models.

Ecosystem

Frontier models sit at the top of the AI capability hierarchy and interact with other models and tools for benchmarking and evaluation.

       ┌─ open models ──┐
input →│ frontier model │→ benchmark
       └─ smaller models ─┘

Misconceptions

MisconceptionRebuttal
Frontier models are always openThey are usually proprietary
Only size defines a frontier modelCapability, not just size, defines them
Frontier models replace smaller modelsThey benchmark and complement smaller models

Trade-offs

  • Performance — often proprietary and expensive
  • Benchmarking — may not be task-optimized
  • Innovation — can stifle open model development

Seen in