Methodology
Every model is scored on a normalized scale from 0.00 to 0.99. A model's score is calculated from a set of underlying factors and then ranked relative to every other model in the field.
What goes into a score
Scores are derived from several input factors, including:
- Use-case performance (coding, writing, reasoning, vision)
- Price (cost per token — cheaper is weighted as better value)
- Speed and latency
- Reliability and security posture
- Context window size
These raw factor values are inputs to the calculation only — they are not displayed individually. Only the resulting score is shown.
Segments
The same factors are weighted differently depending on who the ranking is for:
Overall
A balanced blend of capability, reliability, speed and value.
Enterprise
Weighted toward reliability, security, deep reasoning and large context.
Consumer
Weighted toward value for money, speed and everyday usefulness.