Compare Mode

Compare Mode runs the same request through multiple available model routes so you can compare quality, structure, and usefulness.

Before you begin

• Write one prompt that all selected models can answer fairly.
• Attach only the files required for the comparison.
• Decide what makes one output better than another before you run the request.
• Use fewer outputs when speed or token use matters.

Use Compare Mode for

• Important customer-facing wording.
• Policy or procedure drafts.
• Complex technical explanations.
• Script review or documentation alternatives.

Run a comparison

1. Open chat, an agent task, or a workflow step that supports multiple outputs.
2. Set the output count, such as x2, x3, or x4.
3. Choose the model route for each output slot.
4. Submit the same request once.
5. Review each output side by side.
6. Select the output you want to keep.
7. Continue in chat, refine, or export the selected output.

Expected result: the selected output remains available, and you can show all outputs again when you need to compare the alternatives.

Review criteria

Select an output

Choose the response you want to continue with, then refine, export, or continue in chat.

Tip: If all outputs miss the mark, improve the prompt and rerun with fewer model routes before increasing the comparison size.

Tips

• Keep names, prompts, and configuration values specific to the task you are performing.
• Check role, subscription, region, and tenant policy when a feature is not visible.

Troubleshooting

Next steps

• Models and AI overview
• Create agent tasks