Compare

Two models against each other: how noisy each one is, what an answer costs, how much the model writes to produce it, and how the five measures behind the noise score compare one by one.

Every chart below reads the same way: A is the model on the left, B the one on the right. The two often come from the same lab, so the pair takes these two colours rather than their vendor hues.

Head to head

The three headline numbers. Each pair of bars is drawn against the same scale, and the grey tick is where the field of twenty models averages out.

The five measures behind the score

The noise index is the simple average of these five, so this is where a gap in the headline number comes from. Each row is one measure on its own axis-wide scale; the bar between the two dots is the gap, and the tick is the field average for that measure. What each one asks is on the Details page.