Google announced on Thursday (4) the launch of Gemini 3 Deep Think, its most powerful artificial intelligence model to date. The new mode introduces major advances across several capabilities, but it is available exclusively to subscribers of the company’s paid AI plans.
According to Google, Deep Think delivers significant improvements in reasoning, complex mathematical and logical problem-solving, and scientific analysis. The company emphasizes that the model’s performance has been validated through some of the industry’s most demanding benchmark tests.
In its announcement, Google highlighted:
“Gemini 3 Deep Think is the market leader in rigorous benchmarks such as Humanity’s Last Exam (41.0% without the use of tools) and ARC-AGI-2 (a record 45.1% with code execution).”
Deep Think is built on advanced variants of Gemini 2.5 Deep Think. Its strong performance comes from its ability to evaluate multiple solutions and reasoning paths simultaneously, without needing to complete one line of thought before exploring another — a trait that allows deeper, more flexible problem-solving.
Who can access Gemini 3 Deep Think?
The new mode is available only to subscribers of the Google AI Ultra plan, which costs US$249.99 per month (with a promotional 50% discount on the first three months)
Using the feature is simple: subscribers just need to choose “Deep Think” from the mode selector in the Gemini prompt.
 the launch of Gemini 3 Deep Think, its most powerful artificial intelligence model to date.){kind=link}