Choosing the right model
Each chat runs on a Model tierA quality level for the AI model behind a chat: Everyday for fast, lightweight tasks, Advanced for challenging multi-step work, and Strategic for the most complex jobs. Higher tiers reason more deeply; lower tiers respond faster and use fewer credits.. Open Choose AI Model from the chat's controls and pick the tier that fits the job; the choice sticks for that conversation, and you can change it mid-chat.
The three tiers
Fast, efficient responses for routine tasks and quick interactions. Summaries, quick lookups, rewording.
Powerful reasoning for complex analysis and detailed work. Questions that have to hold several threads at once.
Premium capability for critical decisions and complex problem-solving. Deep reasoning, intricate data questions, building things.
The tier names stay stable even as the underlying AI models improve, so "Advanced" is always the current advanced model, and a chat you set up once keeps getting better without you touching it.
Your organization may also expose models outside the tiers under Other Models. The dialog is searchable when the list gets long; type a model's name or part of its description.
Why not always use Strategic?
Higher tiers reason more deeply and take longer, and they spend more of your organization's AI credits per answer. Matching the tier to the task is the practical move:
- "Reword this paragraph for a customer email": Everyday answers in seconds.
- "Compare this quarter's returns by region and explain the outliers": Advanced is built for it.
- "Design a scoring model for at-risk accounts from these three Datasets": Strategic.
Reasoning effort
Within a tier, a chat can also set how hard the model thinks before it answers. Open the chat's Settings & Tools menu, choose Reasoning effort, and pick a level (a flyout on the desktop, a row of buttons in the sheet on a phone):
- Auto leaves the decision to Informer: the model's own default, or the setting your administrator chose for the App Builder.
- Low, Medium, High and Max pin the level for that chat. Lower levels answer faster and spend fewer credits; higher levels reason longer on hard problems.
The choice sticks for the conversation, like the tier, and you can change it at any point. A routine follow-up in a Strategic chat is a good moment to drop to Low; a tricky design question is a good moment to raise it. The section appears only when the chat's model accepts an effort level. Informer knows which models do; an administrator can override that per model under AI & Billing.
Seeing what a chat costs
Spend stays out of the way until you ask for it. Open the chat's action menu (the three dots) and choose Show spend: a chip appears beside the menu with what the conversation has cost so far, and clicking it opens the split by billing class, by activity and per model call. It applies to that chat for as long as you stay in it. Open a chat again and spend is hidden until you ask, and Hide spend puts it away sooner.
Amounts are in credits, or in dollars when the chat ran on your own provider key. In the App Builder the chip sits in the preview toolbar instead, titled Build cost, with a history of every build of that App. The same figures stay available after you save, under the App's Usage tab.
Who can use which tier
Which tiers you can pick is set per user by your administrator, through usage profiles under AI & Billing. A tier outside your profile shows a lock with a note naming who can help: "Your AI & Billing admin can grant access to Strategic models." Ask your administrator if you need it.
Some surfaces fix the model deliberately. The App builder always uses Strategic, because app-building involves long, multi-step plans and code generation that lower tiers degrade, so it doesn't offer a model picker at all.
Troubleshooting
- Answers feel slow. Check the tier: Strategic deliberately takes its time. If the question is routine, switch the chat to Everyday.
- Answers feel shallow or lose the thread. The opposite trade: move up a tier and re-ask.
- A tier is locked. Your usage profile doesn't include it. The lock's caption names who can grant access (your AI & Billing admin).
- "No models available." Your administrator hasn't enabled any models for your profile yet.
- "AI is not available." No AI models are configured on your server at all; an administrator needs to activate the license and set up at least one AI provider.