Gemini 3 Deep Think’s advanced capabilities now available for Google AI Ultra subscribers

Google has officially started launching its new updated artificial intelligence (AI) model Gemini 3 Deep Think which is now available from the Gemini app for subscribers of Google AI Ultra who will be able to test their abilities to solve more complex problems.

The company unveiled the progress of its latest deep reasoning model last November, when it detailed that includes enhanced multimodal reasoning and understanding capabilities in order to offer better answers when addressing complex math, science and logic problems.

Now, Google has announced that Gemini 3 Deep Think is now available for users subscribed to Google AI Ultra globally who will be able to begin using their advances to solve complex problems from the Gemini app.

This is what he detailed in a statement on his blog, where he stressed that it is a “industry leading” model, with high scores in well-known ‘benchmarks’ such as Humanity’s Last Exam, where it obtains a 41 percent without using tools, and the ARC-AGI-2, where it achieves an unprecedented 45.1 percent with code execution.

These results are obtained, in part, thanks to the fact that it has been designed to use a advanced parallel reasoning system, that explores multiple hypotheses simultaneously, based on variants of the predecessor model, Gemini 2.5 Deep Think.

To access Gemini 3 Deep Think mode, simply select the ‘deep reasoning’ option in Gemini app prompt baras well as the Gemini 3 model in the models drop-down menu.