Model Guide

Should you commit to a model today? Subscriptions are not hardware: a successor upgrades you in place, so our badges score access and commitment risk, not release age. Adopt Now = GA, fully available, stable. Proven = late-cycle and battle-tested, often the best value here. Rolling Out = new, but gated, staged or in preview. Wait = a successor is close enough that adopting now means migrating twice. Legacy = the line has ended or a deprecation is set. The cycle charts still matter: they predict the successor windows behind these calls.

Cycle bars show days between flagship releases, computed from the dated releases listed on each card (older dates approximate). Current-cycle bars are highlighted and count days since the current model shipped, as of .

Roundups: Everything We Know

One living page per upcoming model, aggregating every credible rumor, leak and report, with rumor scores and source reliability, so you can weigh them yourself.

Benchmarks

Real, independent third-party measurements first: the headline score is the Artificial Analysis Intelligence Index (or LMArena Elo where no AA index exists), ranked AA → Elo → our editorial score. Gaps are shown as -, never filled in. Our own category scores live in the separate editorial assessment below, see methodology.

Leaderboard, real scores

Compare All Models

Click a column header to sort. Click a model name for its full product page.

Reviews

Our verdict on each current flagship: what it is best for, and where it falls short. Full scores live in Benchmarks.

Top 3 at a Glance

API Pricing

Review Verdicts

Live Wire

Real headlines from lab blogs, AI news outlets and research feeds, newest first. Titles link out to the original article.