Skip to main content
The slug (technical identifier) is the stable value Tess uses to reference an AI model in programmatic configuration — for example, when importing a Digital Employee via JSON in Cowork. It is different from the display name (what you see in the UI, such as “Claude 4.5 Haiku”). In JSON, you use the slug (claude-4.5-haiku).
For per-model costs and other slugs, see Models and Costs.

​
Why this matters in Cowork

When you build or import a Digital Employee template pack (JSON), the field that sets the model is model_override. The importer already accepts this identifier: if the value is a model allowed in the target agent’s configuration, Tess applies the override correctly. What was missing was publishing these identifiers clearly in the documentation.

​
How to use it in Cowork JSON

In the pack (or Autopilot / Digital Employee configuration snippet), set the slug in model_override:
The slug must exist in the target agent’s model list (model / agent_model options). If the value is not available for that agent, the import warns and the override is not applied.
Accepted aliases in the pack: model_override or, in some exports, the model field with the same meaning.

​
How to find the correct slug

Use any of these paths:
  1. Model picker in the UI — the technical value saved on the employee / agent is the slug (same pattern as the table below).
  2. Export a pack from a Digital Employee that is already configured — the exported JSON includes model_override with the current slug.
  3. Reference table below — confirmed slugs for common text models.

​
Reference table (text models)

Availability depends on your plan, team/workspace policies, and the models enabled on the agent. Newer models follow the same slug pattern (lowercase, hyphens, and dots for the version — for example claude-4.5-haiku).

Detailed configuration

Model field in Advanced settings

Models and Costs

Lowest possible execution cost per model