Auto: let Multilo pick the model

Different questions deserve different models. Rephrasing a sentence is a job for a fast, inexpensive model; checking whether a source really supports a claim, or synthesizing ten papers, is worth the most capable one. Auto makes that choice for you, message by message, so you get good answers without thinking about model names — and without leaving strong work on a weak model.

What Auto does

When Auto is on, Multilo looks at what each message actually involves — how much text and context it carries, whether it includes an image or a PDF, whether it touches your sources or citations, and how demanding the request is — and answers it with the model that fits. Simple asks go to a fast model; harder or accuracy-critical ones go to a stronger model.

The guiding rule is research-first: when a message involves your sources or citations, Auto leans toward a more capable model rather than the cheapest one, because a wrong answer costs a researcher far more than a few credits.

Turning on Auto

Open the model menu at the bottom of the chat and choose Autoat the top of the list. That’s it — every message from then on is routed for you. New accounts start on Auto by default. You can switch back to a specific model any time from the same menu.

Seeing which model answered

Auto is never a black box. Every answer it routes carries a small “Answered by…” line, and hovering it tells you whythat model was chosen (for example, “citations in play → accuracy-first” or “simple ask → fast model”). Over time this also teaches you which kinds of work benefit from which models.

Retry with a stronger model

If an answer underwhelms, look for the Retry with a stronger modeloption next to the “Answered by” line (it appears when a more capable model is available to you). One click re-runs that same message on the stronger model — and you stay on Auto for everything after.

Which models did I use?

Type /modelsin the chat to see a summary of every model Auto used in the current session, with the reason for each. It’s a handy record for reproducibility — you can note exactly which models contributed to a piece of work.

It never leaves you stuck

If your credit balance runs low, Auto quietly shifts to a lighter model so you can keep working instead of hitting a wall. And if a request would genuinely benefit from a stronger model than your plan includes, Auto uses the best model you do have and lets you know a stronger one is available — it never blocks you.

When to pick a model yourself

Auto is the right default for almost everything. Pick a specific model by hand when you want a run to be perfectly consistent (the same model every time), when you’re deliberately comparing two models on the same task, or when you simply prefer one. Choosing a model manually turns Auto off for the chat until you switch it back on. If you want to understand the tiers behind these choices, see Choosing a model for your task.