Deep Research
Thinking mode locks in a deeper scan (at least 6 steps and 7 links) so the agent can think longer.