Compare model responses

After you create a case, every selected model appears in its own panel in the comparison workspace.

Compare the answers

Look across the panels for agreement, important differences, omitted findings, and differences in reasoning or report structure. A confident answer from one model is not a substitute for clinical review.

RadAIChat also creates an automatic comparison that groups agreed, disagreed, and unique claims. Select a claim to highlight its supporting source text in the model panels. See Understand comparison claims and evidence for details and limitations.

Ask a follow-up question

Write a follow-up in the message box to continue working with the models. Check which models will receive the message before sending it.

To address one model only, expand its panel before sending. Read Use single-model chats and message actions for desktop and mobile instructions.

If a model is unavailable or does not support a setting such as enforced zero data retention, RadAIChat displays a warning so you can adjust the selection or settings.

If a model response fails, you can retry the original prompt or edit it before resending to one or all failed models. See Retry or edit failed messages.

Add context carefully

Only add information that is necessary for the task. Follow your organization’s privacy requirements and review RadAIChat’s Privacy Policy before working with sensitive data.