The living observatory of global AI model releases — frontier systems & open-weight breakthroughs, tracked continuously.
Get the AppFrontier models are landing faster than the papers describing them can be read, and open-weight systems are closing on the leaders. This observatory tracks that acceleration as it happens.
Composite 0–100 scores — the per-category median across independent sources (benchlm.ai, LMArena, Epoch AI) where two or more cover a model, single-source and estimated scores labeled as such in each model's detail view. Pick a category to see which models lead.
A notification the second a new model is reconciled — not a page you have to remember to check.
Tap the days-ago tile to open exactly what shipped in the newest release, then straight into any model.
Opens instantly, keeps the last dataset cached, and refreshes itself whenever you come back to it.
Same API, same benchmarks, same origin flags — built for a phone instead of squeezed onto one.