FAQ

Is it really free?

Yes, and there is no paid tier to buy. It runs on the free tiers of several inference providers, which is why there are message limits rather than a bill.

Which model is answering me?

That is deliberately not shown, anywhere outside the administrator's own panel.

It is not evasion, it is the product being honest about a moving target: your message may be answered by any of several engines depending on which is fastest and healthiest at that moment, and it can change mid-conversation. A model name in the interface would be a label that is true for about a second, and it would invite you to make decisions on the basis of it.

What you are shown is real: the mode you chose, and whether the assistant treated your message as a quick question or a detailed piece of work.

Why did an answer restart?

The engine answering you failed partway through. Rather than continue with a different one and glue two half-answers together into nonsense, the assistant discarded what it had and started over. You get one whole answer instead of a seam in the middle.

Why was an answer slow, and then a different one instant?

Free-tier models are inconsistent — the same model can answer in a third of a second one day and take nine seconds the next. Unizune AI starts the fastest candidates in parallel and takes whichever speaks first, so most of that variance never reaches you. When everything is having a bad moment, it gives up rather than leaving you watching a blank screen, and says so.

What are the limits?

Per minute, per day and per month, plus a separate daily cap on generated images. The My usage tab shows your position against them. They exist to protect a shared quota, not to upsell you — there is nothing to upsell to.

Can it see images?

Yes — attach one in a normal chat. Messages with attachments are routed only to engines that can actually process images, so an attachment is never silently dropped and answered around.

Generating images is a separate tab; see Image generation.

Can it read a PDF or a document?

Yes — attach a PDF or a text file in a normal chat. See Files and documents for what gets extracted, the limits, and what happens with a scanned PDF.

Can it search the web?

Not on its own. It can use tools you connect to it — see Connectors — but there is no built-in browsing, and it will not pretend to have looked something up.

Where is video generation?

Not built. Free-tier video models are currently slow enough and poor enough that shipping one would make the product worse rather than better. The Video tab collects an email address so the demand is measurable, and that is all it does — no account needed, and we will not use the address for anything else.

Does it remember previous conversations?

No. Each chat carries its own history and nothing crosses between them. If you want it to know something every time, put it in a persona.

Is my conversation private?

A chat is private to your account until you explicitly create a public link, and turning that link off stops it resolving. Prompt content does reach third-party inference providers, which the Privacy Policy sets out in full.

Something is broken. Now what?

Send it again first — on free tiers a failure is often momentary. If it persists, the contact address in the footer reaches someone who can look at the logs, and a rough time helps.