Choose an expert and action, enter your request, then hit Run.
Results appear here with token count and latency.