Set the Thinking Directive for a Tier

This guide explains how to set the Thinking directive for a tier on the Tier Map of a PaletteAI Inference Launchpad appliance. For what the directive does and how each mode is interpreted, refer to The Thinking Directive. For the value each mode carries at request time, refer to Thinking Directive Modes.

Prerequisites

• A running PaletteAI Inference Launchpad appliance, with the console reachable.

• Console access with permission to edit routing settings.

• At least one tier with a resolved model on the Tier map. To place a model on the appliance, refer to Deploy a Model.

Set the Thinking Directive

1. From the left main menu, select Routing.

2. On the Tier map card, find the row for the tier you want to change and select edit.

3. In the Thinking selector, choose off, on, or budget.

4. (budget only) In the field next to the selector, enter a positive token count, such as 4096.

5. Select Apply tier and confirm the plan card.

Validate

• The Tier map row shows the new directive in the Thinking column, formatted as on, off, or budget:<n>.

• Send a request through the tier and open its trace. On a reasoning-capable model, the response reports a non-zero reasoning-token count.

Next Steps

• The Thinking Directive

• Thinking Directive Modes

• Manage a Client's Model Access