Every model, compared on what's checkable.

The ones marked "In app". The table opens filtered to them, matched by name against the StickyPrompts model selector on 29 Sept 2026. The rest of the list is the wider market, kept here so you can compare specs before you ask us for a model. Your admins can narrow the list further with Model & Provider Access, so your own workspace shows the intersection of the two.

Because we would have to make it up. A quality index is a measurement: somebody has to run the evals, publish the methodology and stand behind the numbers. Artificial Analysis and ModelBeacon do exactly that, and they are the right places to look. What we can tell you honestly is what each model costs, how much context it holds and what it accepts, with a date on it.

They were synced on 29 Sept 2026 from the OpenRouter public model catalogue, and that date is the real one, not a rolling "updated today". Providers change prices and retire checkpoints with little notice, so treat this as a shortlist tool and check the provider's own documentation before you sign anything.

The table comes from a public aggregator, and a few models we serve are not listed there, or are listed under a different name. Image and video models are not in it at all: Veo, Kling, Seedance, Runway and FLUX live on the image and video pages instead. The model selector inside the app is always the final word.

Yes. Admins allow or block each provider in Model & Provider Access, or limit it to specific teams. If you require EU-served inference or zero data retention, models that cannot meet it are greyed out of the catalogue automatically.