• Native iPad layout: a larger, balanced split view with sidebar, model list and detail side by side, built for iPad instead of a stretched iPhone screen. • Refreshed colors and design throughout. • Faster loading. • Release dates now show correctly in every time zone. • Performance metrics now appear in a clearer order: Cost, Context & memory, Speed, Under load, Optimizations. • Labs without a name now sort last, so a real lab always leads the list.

Benchmark Trends: model names sit on their own full-width line above the score bars so long titles are no longer clipped. Ranking credibility: Trends require at least two sources, quarantine obliterated/junk names, show Median · N provenance chips, and keep Grok 4.7 on the coding board where it belongs.

New app icon — a singularity event horizon with a blue-to-orange accretion rim and solid metallic MT letters. The rest of the observatory is unchanged: live model drops, efficiency ranking, clearer Price and Context, and a release-trend chart you can expand.

New-model alerts are more reliable. Invalid device tokens are dropped so pushes keep reaching the phones that can still receive them.

Refreshed app icon — the same MT mark now on your Home Screen and on modeltracker.ai. Efficiency ranking stays in Benchmark Trends: find models that score well and cost less per token. Clearer Price and Context — tap either pill on a model’s detail card. Price explains input vs. output cost per million tokens. Context explains the window size — working memory for one request, not parameter count. Where scores come from — each model’s Benchmarks heading shows whether numbers are a median across independent sources, reported from the model’s own materials, or estimated. The release-trend chart expands to a full-screen view with monthly volume on its own axis under the running total. Landscape rendering is corrected. Cadence uses a tighter window so it tracks the current release pace.

Efficiency ranking — a new Benchmark Trends category for models that score well and cost less per token, so you can pick a leader that is also cheap to run. Clearer Price and Context — tap either pill on a model’s detail card. Price explains the two dollar amounts (input vs. output, per million tokens). Context explains the window size — working memory for one request, not parameter count. Params, when listed, explains what a parameter is. Where scores come from — each model’s Benchmarks heading now says whether the numbers are a median across independent sources, reported from the model’s own materials, or estimated. The release-trend chart expands. Tap it for a full-screen view where each month’s volume sits on its own axis beneath the running total, so you can see both how far the curve has come and how busy any single month was. Rotating to landscape now renders that view correctly. The cadence figure uses a tighter window, so it tracks the current pace rather than averaging it away.

More accurate counts: the tracked-model total now reflects every release the observatory has ever seen, and keeps climbing as new models land instead of plateauing. A truer release trend: past months keep every model that shipped in them, so the curve reflects what actually happened rather than only what's currently listed. Better country labels: models from smaller labs and re-publishers now resolve to a flag instead of an unknown globe, with the United Arab Emirates added to the list.

New home-screen widget: the latest drop and the top of the leaderboard, live, without opening the app. A new "Why this exists" card tracks the release cadence and the model of the moment — tap the featured model to jump straight to it. A living release-trend chart: every model release since ChatGPT, now with a clear axis for both the running total and month-by-month volume, plus milestones that annotate themselves as new leaders land. Plus sharper toolbar contrast and smarter unlock handling throughout.

Version 1.26 brings the app and the website together, plus some sharper tools for digging through the catalog. Unlock the web, too — your purchase now works at modeltracker.ai as well. Tap the "Unlock Web" button, sign in with Apple, and your full app experience — search, filters, rankings — carries over to your browser. Once you're unlocked, the button gets out of the way entirely. Smarter search — results now match on a model's country or lab of origin, not just its name. New "This Month" filter — jump straight to everything released so far this month. Clearer leaderboard comparisons — the leading score always appears on the left, so head-to-head comparisons read at a glance. General reliability and performance improvements under the hood. As always, this is a one-time $2.99 purchase — no subscriptions, no ads, no in-app purchases.