Google's new Deep Think feature is here — what it does and why it might not stay Ultra-exclusive for long
Starting today, Deep Think is available to Google AI Ultra subscribers, with a broader API release for trusted testers expected in the coming weeks.
Google says Deep Think excels at challenges that require multi-step reasoning, strategy and iteration. Think of it as giving Gemini more thinking time before the AI responds.
Now, instead of rushing to deliver an answer, Gemini engages in parallel thinking, a process that allows it to generate multiple ideas, evaluate them simultaneously and refine its approach before responding.
It's a move that brings Gemini one step closer to emulating human-style problem-solving, whether it's debugging code, website building or tackling advanced math.
Internally, the model has reached bronze-level performance on the 2025 International Mathematical Olympiad (IMO) benchmark and outperformed competitors on difficult reasoning tests like LiveCodeBench V6 and Humanity's Last Exam, which measure capabilities in science, coding and complex logic.
The company also tested Deep Think with academic researchers, including mathematician Michel van Garrel, to explore its potential for tackling theoretical problems and advancing scientific discovery.
Get instant access to breaking news, the hottest reviews, great deals and helpful tips.
While the original research model behind Deep Think took hours to solve Olympiad-level math problems, the version launching today is optimized for faster, more practical day-to-day use.
To try it out, Ultra subscribers can simply toggle Deep Think on in the Gemini app when using Gemini 2.5 Pro.
The mode supports longer, more detailed responses and works with tools like Google Search and code execution, though usage is currently limited to a fixed number of prompts per day.
As with previous Gemini updates, Google is placing a strong emphasis on safety.
The company reports that Deep Think shows improved tone and content moderation over Gemini 2.5 Pro, though it may sometimes refuse benign requests.
Google says it's actively performing frontier safety evaluations and working on mitigations as model capabilities advance.
For Ultra subscribers, this new feature should support app building, exploring a scientific theory and many more tasks that require a more thoughtful response.
While Deep Think is currently limited to $249/month Gemini Ultra subscribers, there's a good chance it could become more widely accessible in the future. Google has a history of offering some of its most powerful AI features, like Gemini Pro and image generation, to free-tier users after initial testing.
If Deep Think proves successful and stable, it's likely that Google will broaden access to showcase its capabilities and stay competitive alongside other tech giants.
Follow Tom's Guide on Google News to get our up-to-date news, how-tos, and reviews in your feeds. Make sure to click the Follow button.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Forbes
an hour ago
- Forbes
Google's Pixel 10 Pro Pricing Hides Its Subtle Influence
The upcoming 'Made By Google' event on Aug. 20 will see the launch of the Pixel 10 and Pixel 10 Pro families. Recent leaks point to an aggressive pricing strategy that consumers will welcome. Yet Google can extract more value out of the Pixel 10 family than the sticker price. Pixel 10 And Pixel 10 Pro Pricing First up, the current expectation for US prices: The four main models will be priced similarly to the Pixel 9 family. Specifically, the Pixel 10 will start at $799 for the 128 GB model, the Pixel 10 Pro starts at $999, the Pixel 10 Pro XL at $1,199 for 256 GB and the Pixel 10 Pro Fold at $1,799 for 256 GB. Samsung bumped up the prices of the Galaxy Z Fold 7 over the Z Fold 6, and Apple is expected to do the same with the various iPhone 17 Pro models compared to last year's 16 Pro portfolio. Google has decided to stick with the pricing. This will make the new handsets look more attractive to consumers in a head-to-head comparison with the competition. It also offers a certain level of convenience for those looking at the year-on-year evolution. Yet Google can find more value in the package, value that looks to have been passed on to the consumer Pixel 10 Pro BLOCK OF THREE First of all, the Pixel 10 family can shape the smartphone market. The obvious point here is the aforementioned price point. With the Pixel 10 family staying at the lower price point, it pushed other manufacturers to keep their similarly specced handsets at the same level if they want to continue looking competitive. Then you have the tone it sets for the hardware that Google would like to see at each price point. The user experience of the Pixel 10 is set to be enhanced with the addition of a third lens in the main camera. The 'base model' of premium brands has typically launched with two lenses; Google is clearly looking to redefine that part of the spec sheet through the Pixel 10 and leverage it across the Android ecosystem. That will benefit Android as a whole, and anything that benefits Android benefits Google. Finally, you have AI, and this is perhaps the most tangible benefit for Google. Artificial intelligence is one of the key features reshaping not just smartphones but the entire digital ecosystem. Google has been leaning heavily into its AI-powered assistant, Gemini, while it continues to add more AI-based features into its own software stack and Android. Again, this adoption pushes the market towards Google's vision of artificial intelligence on a mobile device. By having a first-mover advantage and establishing this as the way forward, it makes it harder for other platforms and manufacturers, such as Apple's Intelligence on the iPhone. As for the tangible, to make the best use of Gemini and Google's AI suite involves a subscription to Google AI Pro or Google AI Ultra. Trial memberships to these services are available with the purchase of a Pixel device, creating an easy on-ramp to an ongoing subscription, which in turn pulls the consumer even closer into Google's cloud system while pushing back others. The Google Advantage In The Pixel 10 And Pixel 10 Pro The Pixel 10 and Pixel 10 Pro will undoubtedly garner critical acclaim and continue the commercial success of the previous Pixel handsets. Additionally, Google will use these handsets to maintain premium pricing across the smartphone market, promote an increase in specifications at each price point, and bring more users into Google's vision and implementation of artificial intelligence. Google is expected to launch the Pixel 10, Pixel 10 Pro, Pixel 10 Pro XL, and Pixel 10 Pro Fold on Aug. 20, with retail sales to follow the following week. Now read the latest Pixel 10 Pro, Samsung Galaxy, and Android headlines in Forbes' weekly smartphone news digest...
Yahoo
an hour ago
- Yahoo
Google Has Quietly Been Detecting Earthquakes by Sensing Rumbling in Android Phones For Years
Google has for years been harnessing the power of its Android smartphones to detect and measure tens of thousands of earthquakes. In a new paper published in the journal Science, researchers from the search giant described how they used motion sensors from its two billion-strong network of phones running Android between the years 2021 and 2024 to detect and alert quakes to users in almost 100 countries around the world. Known as "Android Earthquake Alerts" (AEA), this early warning system has uses the smartphones' accelerometers to detect telltale vibrations as they happen and inform residents of quakes in their areas. With its two-tiered severity levels — the weaker "be aware alert" for lower-magnitude quakes that sends typical push notifications, and the more pressing "take action alert" for moderate or extreme shakes that issues a loud beeping alarm that overrides "Do Not Disturb" settings — Google boasts that it has warned of more than 11,000 quakes. While such a system has invariably saved lives, however, AEA is far from flawless. In the case of Turkey's deadly quakes in 2023 that claimed more than 55,000 lives and injured over 100,000 people, the company now admits that it erroneously sent its lower-level notifications to hundreds of thousands of people — a massive failure during one of the deadliest natural disasters in recent history. Both in the Science paper and in a statement to BBC, Google admitted that it only sent 469 "Take Action" alerts during the first of the two earthquakes that struck Turkey and Syria in February 2023, and instead mass-notified half a million people with the "Be Aware" version, the less severe of the two warnings. Because the first earthquake occurred at around 4:15 in the morning, that quieter warning likely went unseen by people who were asleep and had their phones silenced. That discrepancy appears to be the result of bad seismic readings from the phones' sensors. Initially, the system's algorithms estimated that the first quake was between a 4.5 and 4.9 on the moment magnitude scale (MMS), but it was actually a far more serious 7.8. During the second quake, the AEA also underestimated the magnitude, sending only 8,158 "Take Action" alarms and almost four million "Be Aware" alerts. (That shoddy algorithm, the company says, has since been updated.) Until now, Google deflected blame for the system's failure, even after the BBC investigated the AEA in the aftermath of the quakes and found it to be severely lacking. Though this sort of admission is better late than never, it's still concerning that Google took more than two years to admit the system's failures during the Turkey earthquakes. With so many people relying on the tech monolith for life-saving alerts, taking accountability for the AEA's failures is paramount — and that accountability should have been much swifter. More on disasters: Extreme Heat Is Killing Unfathomable Numbers of People Worldwide Solve the daily Crossword
The Hill
2 hours ago
- The Hill
Electricity cost concerns grow amid push for more AI data centers
As tech companies race to power data centers to support artificial intelligence services, some Americans are reporting spikes in their electricity bills. The Trump administration recently released an ' AI Action Plan,' which vowed to expedite permitting for the construction of energy-intensive data centers. During the unveiling of the plan, President Trump acknowledged the growing need for power to run these centers, which are being built across the country by companies like Google, Amazon and Microsoft. New reporting by the Washington Post suggests consumers are seeing a spike in their energy bills because of these data centers. According to the report, customers in the areas serviced by PJM Interconnection have seen their electricity bills increase. In Ohio, prices are up between 10 percent and 15 percent, while in New Jersey, prices have increased 20 percent. Experts have told NewsNation that a 5-gigawatt data center could use more energy a year than many states — but they insist Big Tech growth isn't solely to blame for the recent price increases. 'Data centers are affecting electricity, but they are affecting electricity planning at this point. They are not necessarily affecting electricity bills,' said Tricia Pridemore, president of the National Association of Regulatory Commissioners (NARC). 'These large AI data centers that you keep hearing about on the news, they are in the planning and construction phase at this point.' Supply and demand are believed to be behind cost increases in the PJM territory, which has a reorganized market. They have a capacity auction in which power industries bid to supply enough electricity in the future. Right now, there is a greater demand because more supply will be needed for these data centers. At PJM's recent auction, prices went up a record 22 percent, according to Reuters. 'Within reorganized markets such as PJM, MISO, SPP, they go out and use independent power producers to generate electricity,' Pridemore said. 'Based upon the electricity that is available, those market prices drive the cost of the electricity to anyone in the system, whether it's a residential customer or a commercial or a business customer.' The NARC president added that data centers can reduce their power usage and cost by using renewable energy, and some are already tapping into solar energy.
