How FrontierDot labels benchmarks
Every benchmark result is labeled with one evidence type: provider_reported when the model provider published it, third_party_reported when an outside evaluator published it, or measured_by_frontierdot when FrontierDot ran it.
Those labels stay separate. Provider-reported numbers and FrontierDot measurements are not mixed into one unlabeled table, and this site does not name a universal model winner.
A FrontierDot-measured run records the exact model, the run date, the harness version, the settings, the scoring method, and the limitations.
Tool comparisons are separate from benchmark scores. They take pricing, free tier, and features from each tool's own site.
How to choose an online file converter (not a benchmark score)
How comparisons are chosen English tool comparisons
FAQ
- What is a benchmark on FrontierDot?
- A benchmark result on FrontierDot is labeled with one evidence type: provider_reported when the model provider published it, third_party_reported when an outside evaluator published it, or measured_by_frontierdot when FrontierDot ran it. Those labels stay separate, and this site does not name a universal model winner. Tool comparisons are separate from benchmark scores.
- Does FrontierDot publish a mapping benchmark?
- This page does not publish a mapping-benchmark score. A measured run is stored only when FrontierDot has run it, under the evidence label measured_by_frontierdot. There is no such score on this page.