Google is expanding access to its newest language model capabilities by rolling out Gemini 3 Deep Think to Artificial Intelligence Ultra subscribers. The premium feature is described as a supercharged mode of the Gemini 3 large language model, and it is aimed at handling more complex and demanding problems than the standard configuration. The rollout follows the late November launch of Gemini 3 and positions Deep Think as the latest evolution in Google’s efforts to compete in the advanced Artificial Intelligence assistant space.
According to Google, Gemini 3 Deep Think “uses advanced parallel reasoning to explore multiple hypotheses simultaneously,” which is meant to improve its performance on intricate tasks that require deeper analysis. The company says the model scored 41 percent on the extremely difficult large language model benchmark called Humanity’s Last Exam, giving Google a specific metric to point to as a bragging right following Gemini 2.5. If you are paying $250 a month for the pro tier Artificial Intelligence Ultra plan, you can access Gemini 3 Deep Think in the Gemini app by opening the prompt bar and selecting the new mode from a dropdown menu.
Google’s broader Artificial Intelligence ecosystem is seeing strong demand, with the company stating that its Gemini app is used by 650 million users every month. The launch of Gemini 3 has coincided with competitive pressure on rivals, with a report saying that OpenAI reportedly lost 6 percent of its users in a week following Gemini 3’s arrival. At the same time, Google has had to manage resource constraints, recently capping access to its next generation image generator Nano Banana by temporarily dropping the number of free image prompts from three to two, although this change did not affect Google Artificial Intelligence Pro or Artificial Intelligence Ultra subscribers. Together, these developments signal Google’s push to differentiate its paid tiers with more powerful tools like Gemini 3 Deep Think while balancing usage limits on free offerings.
