Skip to main content
Every agent runs on a language model. The model decides how deeply the agent can reason, how much context it holds and how much each turn costs. Agent model

​
What the Model screen shows

  • Connected - the model the agent is running on, with figures such as context window and maximum output length.
  • Workspace models - the list of models the workspace allows. Click Use on a card to switch the agent over.
Each card carries capability labels - Reasoning (for multi-step thinking) or Vision (can read images), for example - so you can tell at a glance which model suits the job.

​
Choosing the right one

After switching models, run Test chat through a few familiar scenarios. The same Instructions can read very differently to two different models.

​
How to switch an agent’s model

1

Open the Model section

On the agent page, go to the Agent builder tab. The Configuration column on the right shows the model the agent is running on.
2

Look at the alternatives

Click the Model box for a short list, or click View all models to open the workspace’s full catalogue with figures to compare - context window, output length and capability labels.Model dropdown
3

Click Use

Click Use on the model you want. The agent switches over immediately - there is nothing else to save.Model list
There is no add, remove or enable/disable for models at the agent level. The catalogue is a fixed workspace list (Resources > Models), and an agent always runs exactly one model at a time - so switching is the only action there is.