Signal Breakdown Explorer

Signals Tracked

6

unique quality dimensions

Models with Signal Data

300

of 300 total models

Avg Signals per Model

5.8

signals per model on average

What Signals Represent

Signals are individual quality dimensions that capture different aspects of a model's value. Each signal measures a specific attribute such as benchmark performance, pricing efficiency, context capacity, or capability breadth. Together, they provide a multi-dimensional view of model quality.

How Weights Are Applied

Each signal is assigned a weight reflecting its importance in the overall assessment. Weights are expressed as fractions summing to 1.0 (100%). A signal with weight 0.25 contributes up to 25% of the composite score. Weights are calibrated based on the signal's relevance to practical model quality.

Normalized Score (0–100)

Each signal's raw value is normalized to a 0–100 scale to make signals comparable regardless of their original units. A score of 100 means the model ranks at the top for that signal, while 0 indicates the lowest possible performance. Z-scores are computed first, then mapped to the 0–100 range.

How Contribution Works

A signal's contribution equals its weight multiplied by its normalized score. For example, a signal with weight 0.25 and normalized score 80 contributes 20 points to the composite score. The sum of all contributions gives the final composite score. This makes it easy to see which signals drive each model's ranking.