Model API pages now show what a model expects and returns

Honestly, I thought the playground already solved this. You fill in a form, run a model, see what comes back, flip to JSON if you want the raw shape. It's all there.

That's true, but only if you're on the page. Most people building something are not sitting in the playground. They're writing code somewhere else, and until now there was no way for them to know what a model supported without pulling the model's details themselves, then reading through them by hand.

That's the complaint I kept hearing. Not "let me try this model", people already had that. It was "let me see what this model supports before I touch any code."

What we built

Every model's API page now renders a real reference for it. Field names, types, whether something is required, laid out the way a proper reference should be. What you send the model, and what you get from it, shown the same way, right on the page.

Screenshot showing input and output schemas of meta-llama/prompt-guard-86m

What it feels like

You open a model's page and you already know what it supports. Nothing to run, nothing to fetch, nothing to read through by hand. It's just there before you've written any code.

That changes how you evaluate a model, especially when you're choosing between a few. You read what each one supports, compare them, and decide without running any of them.