Following the launch of Gemini 3 in late November, Google is rolling out its newest Deep Think mode for AI Ultra subscribers.

Now accessible in the Gemini app for users paying $250 monthly for the pro plan, Gemini 3 Deep Think serves as the upgraded mode of Google’s latest large language model (LLM), aimed at tackling more intricate challenges. The company notes that it “utilizes advanced parallel reasoning to investigate multiple hypotheses at once.”

This newest update, succeeding Gemini 2.5, Gemini 3 Deep Think boasts several accomplishments, with Google asserting it achieved a score of 41 percent on the highly demanding LLM benchmark, Humanity’s Last Exam.

If you hold an AI Ultra subscription, you can experiment with Gemini 3 Deep Think in the Gemini app by navigating to the prompt bar and choosing the dropdown option.

There is significant demand for Google’s AI tools lately, even though the company joined competitors such as Anthropic and OpenAI in receiving a mediocre AI safety report card this week. This rollout follows shortly after Google had to temporarily restrict access to its next-generation image generator, Nano Banana, cutting the number of free image prompts from three to two (this did not impact subscribers of Google AI Pro or AI Ultra plans). Google reports that its Gemini app is utilized by 650 million users each month, and with the introduction of Gemini 3, rival OpenAI allegedly lost 6 percent of its user base within a week.