Gemini 3 Deep Think is now available in the Gemini app (blog.google)

Google has announced the rollout of Gemini 3 Deep Think mode for its AI Ultra subscribers, now available in the Gemini app. This new feature offers substantial enhancements in reasoning capabilities, specifically designed to solve intricate math, science, and logic problems that have historically posed challenges for leading AI models. Notably, Gemini 3 Deep Think has achieved remarkable scores on demanding benchmarks, scoring 41.0% on Humanity’s Last Exam and an impressive 45.1% on ARC-AGI-2, underlining its superiority in evaluating reasoning under various conditions. The significance of Gemini 3 Deep Think lies in its advanced parallel reasoning technique, which enables the model to evaluate multiple hypotheses simultaneously. This innovation builds on the success of the earlier Gemini 2.5 variants, which were recognized for their outstanding performance at prestigious competitions like the International Mathematical Olympiad. By allowing Ultra subscribers to engage with this powerful tool, Google is not only advancing the capabilities of AI in tackling complex problems but also pushing the boundaries of what’s achievable in the AI/ML community. Users can access this mode by selecting “Deep Think” in the prompt bar within the app.