Auto
The right model for every message, picked for you
On the Free plan, Babbily runs on Auto. Choosing a model yourself is a paid feature.
How Auto works
A quick step before each reply picks the model for it.
Send your message
Auto reads your latest message along with the recent conversation.
It judges the task
One of 23 kinds of task, from writing and code to math, research, money and law, and whether it’s routine, standard or demanding.
It picks a model
Each kind of task has its own ranking of models. Money, legal and security questions always get a thorough one.
See the models used
Auto · Models used, below the composer, lists every model that has replied in the chat.
Under the hood
How Auto classifies and routes a message
Every message gets its own decision. Auto reads it and scores two things: what kind of task it is, and how demanding. It routes only once it’s at least 70% sure, choosing the lightest difficulty that clears that bar. Each task and difficulty has its own ranking of models and reasoning levels, and your message goes to the top of that list.
23 kinds of task
Conversation, writing, translation, summaries, research, math, science, finance, legal, planning, code, data and more.
3 levels of difficulty
Routine, standard and demanding. Harder work goes to a stronger model with more reasoning, and quick questions stay fast.
When it isn’t sure
If Auto can’t classify a message confidently, or can’t see all of it yet, like an attachment or a project’s files, it takes a dependable everyday route instead of guessing.
Auto in action
Frequently asked questions
Send your first message on Auto
Start a chat, ask anything and see which models answered, right below the composer.
