Google rolls out Gemini 3 Deep Think to Google AI Ultra subscribers in the Gemini app, after saying in November it needed “extra time for safety evaluations”
its most powerful AI model yet Business Today : Gemini 3 gets jailbroken in minutes, teaches how to make smallpox virus: Report Shannon Connellan / Mashable : Gemini 3 Deep Think rolls out to paid subscribers Harsh Shivam / Business Standard : Google rolls out Gemini 3 Deep Think mode to Ultra users: Check what's new Govind Choudhary / Livemint : Google rolls out Gemini 3 Deep Think to deliver advanced reasoning: How powerful is it and who can use it? X: Jeff Dean / @jeffdean : Gemini 3 Deep Think is now available for Ultra users, making available our IMO & ICPC Gold Medal-winning technology. Deep Think shows improved generalization on difficult benchmarks like ARC-AGI-2, and outperforms Gemini 3 Pro on HLE & GPQA Diamond. We hope this serves as a useful step toward ever-more-capable scientific reasoning agents. Demis Hassabis / @demishassabis : Gemini 3 Deep Think is now available for Google AI Ultra subscribers in the @GeminiApp, incorporating our gold medal winning IMO and ICPC technologies! 🏅With its parallel thinking capabilities it can tackle highly complex maths & science problems - enjoy! [image] Noam Shazeer / @noamshazeer : Gemini 3 Deep Think mode is live for Ultra users today. When using parallel lines of thought in this mode, Gemini shows meaningful improvement on key reasoning benchmarks such as ARC-AGI-2 & HLE. I think you will be deeply impressed. https://blog.google/... Quoc Le / @quocleix : Gemini 3 Deep Think mode is now live in the Gemini app for Ultra users. 🚀 Building on the technology that reached a gold-medal level at the ICPC World Finals & IMO, it uses parallel thinking to excel at difficult coding and scientific tasks. https://blog.google/... Sundar Pichai / @sundarpichai : Gemini 3 Deep Think mode brings our strongest reasoning capabilities yet, now rolling out to @GeminiApp Ultra subscribers. Cool example of how it can simulate complex 3D architecture.
Context & Ripple Effects
This is the release phase of a rollout Google had already signaled would wait on additional safety evaluations. It also extends the product pattern established when Gemini 2.5 Deep Think was placed in the $250-per-month Ultra tier.
The launch matters because Google is pairing its strongest claimed reasoning capability with a restricted paid channel, even as outside reporting has raised questions about jailbreak resistance.
First-order effects
- Google AI Ultra subscribers can now use Gemini 3 Deep Think in the Gemini app, making the new reasoning mode an immediate premium-product differentiator.
- Google moves from a safety-delayed announcement to controlled subscriber availability; the reported jailbreak claim makes the quality of its access safeguards newly consequential.
Second-order effects
- Rival frontier-model providers face added pressure to match both advanced reasoning features and the premium packaging used to monetize them.
- The rollout gives Google more real-world use of the model within a bounded audience, while safety performance may influence how quickly it can widen access or offer comparable capabilities through other channels.
Third-order effects
- Frontier reasoning models are increasingly likely to be released through tiered, governed access rather than broad day-one availability, with safety review becoming part of product timing and segmentation.
- If stronger models continue to expose jailbreak weaknesses, competitive advantage will depend not only on benchmark gains but also on whether providers can demonstrate durable controls around high-capability access.
The trend: Frontier AI vendors are turning advanced reasoning into a premium, staged-access product while treating safety evaluation as a core constraint on distribution.