Every model we judge to sit at or near the frontier — the systems a serious team would actually reach for today — filed in one place. Each entry opens a Survey Record: what the model is, what it costs, how it is licensed, and an honest read of where it leads and where it lags.
Placement is our own call, not a vendor’s. Models are grouped by tier — flagship, balanced, efficient, and open-weight — and ordered within the field using independently verified intelligence scores where they exist; figures that are only vendor-reported are flagged as such so you can weigh them accordingly. When a lab ships a direct successor, the older model steps down to Previously Explored at the foot of the page: kept on record and still linked, no longer in service.