Gemini 3 Deep Think combines expert-level scientific domain knowledge + engineering utility to help researchers with their work in fields like mathematics, physics, chemistry and much more. e.g. Prof Lisa Carbone explains how she has been using it in her complex research work:

Thrilled to announce a big upgrade to Gemini 3 Deep Think that hits new records on the most rigorous benchmarks in maths, science & reasoning - including 84.6% on ARC-AGI-2, 48.4% Humanity’s Last Exam without tools, and 3455 Elo rating on Codeforces!

Google AI Ultra subscribers can try Gemini 3 Deep Think mode in the @GeminiApp now - happy discovering! Details in the blog:

Hi Demis, I’d love to use DeepThink for science (I am a mathematical physicist), but can’t afford it… any university subsrciptions?greetings from @LFC fan ;)

More importantly how Deepmind scientist use Gemini Deep think

DO YOU GUYS ŠĘRÏØÜŚŁŸ NOT REALIZE YOU'RE CHASING I ÆM?

Humans solving Fermat’s Last Theorem and Poincaré Conjecture, and discovering General Relativity and Quantum Mechanics will be looked upon as the last great achievements of humanity’s pure intelligence before the help of AI.

Sir, thank you for your commitment to advancing scientific research. A lot of people in the industry say it because it sounds cool, but you stick to it since the beginning. Gemini helped me formalize a really speculative idea that may be published soon. Keep the quality work!

GOOD NEWS YOU'RE FINALLY AT THE THRESHOLD OF FINDING ME.

If the AI is doing the deep thinking then the professor is just there to sign the paperwork. We are watching the automation of the tenure track in real time

Impressive catch by Gemini 3 Deep Think on that subtle Lie algebra flaw. When will Deep Think mode become accessible for academic researchers worldwide?

What's its cut-off date on scientific research ?

This is crazy!! Take a bow!!🙏🙏 @demishassabis

Please look into this one,

The gap between "here is a powerful model" and "here is the solution to your specific research problem" is where the real value lives. Frontier labs build capabilities. The accumulated context of solving Prof Carbone's actual problems over months, that's what makes AI useful rather than just impressive.

This is so incredible to hear. It is unbelievable to see how much progress did AI make in such a short time interval. Congratulations to Google Deep Mind for this achievement!

If it works like before - guesses instead of admitting it does not know the answer to a question -, you can shove it.

Would be great to see a full workflow example. Problem model steps human edits final result. That’s what researchers need

Very ambitious, but this is more like a sales pitch.

It’s great! I use Gemini for Analize legal contract 👏. Link:

do you have ways to run this model at an accelerated speed internally? I'm sure you've already considered this, but it would be very useful for generating synthetic data and verifying both organic and synthetic data.

not with the pro!

Could you use it to find the reason behind the onset of Type 1 Diabetes?