Set how much each requirement matters for your job — reasoning, cost, latency, context, coding, tool use, multimodal — add hard constraints, and get a ranked shortlist of model classes with the reason for each, from an authored, dated matrix. Free. Nothing you set is uploaded.
The authored scores (0–5) behind every result, dated . Qualitative tiers, not live specs — this is the judgement the ranking is built on, shown in full.
The matcher, the ranked shortlist with reasons, and the full matrix are free and stay free. The Architect’s Edition adds what a matrix can’t: when each class is the wrong choice, the cost-control tactics that change the maths, and how to actually evaluate and decide.
Team licence for an engineering org? Email for a group rate.
Unlocks the Architect’s Edition on this device and survives reloads. No account, nothing transmitted.