AI models

The model is the AI that writes an expert's answers. Models differ in speed and in how well they handle long or difficult questions, so you choose which ones each expert may use.

Which models you can choose

The list you see is the models switched on for your plan, grouped under the name of the provider. Because it depends on your plan, and because the platform adds and retires models over time, two companies can see different lists. The names in your own list are the ones to go by.

If your plan includes Provider, you can also connect your own AI key. The models you enable there appear in the same list under your provider's name, and they run on your account with that provider. See Provider: using your own AI key. Provider is a Professional feature.

Choosing models for an expert

There are three places to do it, and they all change the same setting.

• When you create an expert. On the Answering step, tick the models under Models. The first model in the list starts ticked.
• On the expert's Configure page. Open Settings and look at Models. Tick or untick models. Select all ticks every model from one provider. The change saves straight away.
• On the experts list. In All experts, select the button in the Model column to open a small checklist. Use All beside a provider to tick all of its models.

The Model column in the list reads Platform default when nothing is ticked, the model's name when there is one, and a count such as 3 models when there are several. If every model your plan offers is ticked, it reads Any of followed by the number, for example Any of 6 models.

A model that is paused carries , paused after its name. If an expert uses several models and all of them are paused, the label reads, for example, 3 models, paused.

One model or several

• One model ticked. Every answer from that expert uses it. In the chat, the model name is shown beside the message box as plain information.
• Several ticked. The person asking can pick between them from the model button beside the message box in the chat. Each choice shows its name and a short description. Offer a choice when your team has a real reason to compare, for example a quick model for simple questions and a stronger one for long documents.

At least one model has to stay ticked, and the page will not let you untick the last one. If your plan has no models switched on at all, the expert uses whatever the platform default is.

When a model is paused

Sometimes a model is switched off for maintenance. An expert that uses it keeps its choice and shows the model ticked with a Paused label, and you do not need to change anything. Chats use another available model until it is back. You can untick a paused model, but you cannot newly tick one.

A different model for one channel

By default a channel such as your website widget or WhatsApp uses the expert's own model. To use another one for a single channel, open that channel under Integrations, choose the expert, and set Model for this channel. The setting is there on every channel page that offers it, not only the website widget, WhatsApp and Telegram. The first choice, Use expert's model (default), clears the override. The change saves as soon as you pick. See How integrations work.

The channel list offers every model your plan allows, not only the ones ticked for the expert.

Models and templates

An expert made from a template starts on the model that template names, shown on the template page. If that model is not available to you, the expert starts on the platform default. You can change it afterwards in Settings.

What about response effort

Models are the main lever you have over speed and depth. There is no separate effort setting any more. Aimy decides how hard to look on each question.