Two views of model adoption

Developer breadth answers how widely a returned model appears across the cohort. Token share answers how much of the observed workload it handles. A widely tried model can rank below a more concentrated, high-volume model on tokens—and vice versa. Because the endpoint is capped at 60 token-ranked rows, breadth leaders are leaders among returned models, not necessarily every model in the database.

What each adoption metric means

Participating-developer breadth

Distinct listed developers whose eligible usage included a model during the selected window, divided by all listed developers with eligible usage in that window. It is a cohort reach measure, not a count of every developer who uses the model.

Observed token share

A model's input, output, cache-creation, and cache-read tokens divided by all eligible tokens observed across models in the same 30-day cohort window. It measures workload volume, not user preference or model quality.

Normalized model

Raw identifiers that describe the same model can be combined into one display row—for example, provider variants or date-suffixed IDs. The model table reports the normalized key and the aggregate of its variants.

Used through

The coding tools or agents that reported usage for a model, ordered by observed volume. This field provides context but does not imply that the model is exclusive to those tools.