Gemini 2.5 Deep Think Launched for Google AI Ultra Users

Gemini 2.5 Deep Think Launched for Google AI Ultra Users

Gemini 2.5 Deep Think Launched for Google AI Ultra Users

‘Deep Think’ enhances the capabilities of Google’s elite Gemini AI model.

(Image credit: Google)

Essential Information

  • The enhanced reasoning model, Gemini 2.5 Deep Think, is designed to manage complex mathematics, coding, and broad reasoning, now accessible to AI Ultra subscribers.
  • Deep Think ostensibly operates like OpenAI’s leading models, generating intelligent solutions in real-time.
  • It achieved an 84% score on the MMMU assessment and excelled at Olympiad problems, albeit Google withheld certain figures.

Today, Google declared that it’s broadening access to an improved version of Gemini 2.5 Deep Think, now offered to Google AI Ultra subscribers.

During Google I/O in May, the company introduced significant enhancements to its Gemini 2.5 series, which includes Deep Think, a novel reasoning model for Gemini 2.5 Pro tailored to address more demanding tasks such as advanced mathematics and coding.

Following the announcement, Google has been rigorously testing Deep Think. DeepMind CEO Demis Hassabis previously stated that it’s constructed on “the latest state-of-the-art research,” equipping the model with the capacity to consider numerous ideas simultaneously before arriving at a conclusion.