Find your best-fit model

Weights are relative — what matters is their proportion, not the totals.

Intelligence is the Known Good Index. Speed is median output tokens per second across tracked providers. Low cost is blended price, inverted so cheaper scores higher. Each is min-max normalised across every model that has a Known Good Index score, then combined with your weights.

It will not tell you a model is correct for your task. Aggregate scores hide the thing you actually care about, which is usually your own prompts on your own data — treat the shortlist as a shortlist and test the top two or three yourself.

It will not price your workload. We publish blended price per 1M tokens, not cost per task: cost per task needs token counts from an evaluation suite we do not run, so publishing it would mean inventing the inputs. See methodology.

It will not rank a model we cannot score. A model needs at least one science, one mathematics and one coding result before it appears here — 61 of 908 tracked models currently qualify.