Who we are
Unbiased is built by Circuit & Chisel, the team behind Pareto, a blended model that runs several LLMs on every request and synthesizes one answer. The people writing here are the same people running the benchmarks, wiring the meters, and paying the API bills we publish. That is a bias, and we would rather name it than pretend it away: we sell one of the products this blog covers.
How we measure
Numbers on this blog come from three sources, always labeled: metered bills from real API requests we ran ourselves, with the date shown; published list prices from providers' own pricing pages, linked and dated; and same-harness benchmarks from our public evaluation harness, with both sides reported. When a number is arithmetic rather than a measurement, the method is stated next to it. The full methodology lives on the model card.
The editorial rules
Every claim carries a source or a receipt. Prices show the date they were last verified, because pricing content decays. Competitors get their honest best case, including when the honest answer is to use them instead of us. And when we get something wrong, we correct the page and say so rather than quietly editing. If a number looks wrong to you, tell us, or better: run it yourself and publish what you find.
Compare any two models
List rates and dated competitor measurements: prices and measured bills. Pareto 26.9 measured task costs are not published. Verbosity: Edition 2.