Settings - Models
The Models tab is a read-only directory: every model you can point an agent at, grouped by provider, with the capabilities that matter for dashboard work.
Open this dialog
Open Settings from the app menu — Dashboardx Designer → Settings on macOS, Settings on Windows, or press ⌘, (Ctrl+, on Windows) — and select Models under AI & Agent.
To change which model an agent actually uses, go to Settings → Agent. This page tells you what is available.
What each entry shows
Models are grouped under their provider. Only providers you have configured an API key for are usable — see Settings → Providers.
Each model carries up to three tags:
| Tag | Meaning for dashboard work |
|---|---|
| Tools | Can call the tools that build dashboards — reading sheets, creating widgets, binding data. Required for anything beyond chatting. |
| Reasoning | Produces a visible thinking trace. See Thinking & Reasoning. |
| Vision | Accepts images as input, so you can drop in a screenshot of a layout you want. |
Some models also list variants — preset behaviour modes for the same underlying model, selectable in the conversation.
Choosing between the models
The catalog splits into a capable tier and a fast tier, usually named Pro / Max and Flash:
- The capable tier is what you want when an agent is designing a dashboard, choosing charts, or reviewing work.
- The Flash tier is a reasonable choice for short, mechanical edits, and is what the Global Small Model is for.
Every reasoning-capable model in the current catalog also has Tools, so you are not trading capability for thinking.
When a model is unavailable
A model disappears from the selectors when its provider has no key, and it is marked as failed when a key is present but verification failed. Conversations in flight fall back to the model configured in Settings → Agent and say so in the conversation.
Context usage
The ring in the corner of the message box shows how full the current conversation's context window is. It is a reading, not part of the model list. For what it counts, the colour thresholds and the hover breakdown, see Agent Dock.