In a way, Google is making AI reasoning something you can dial in. Whether you’re on the web, an iPhone or an Android in India or elsewhere, you’ll find an Extended Thinking mode to go with the quicker Standard one. It’s a more assertive way of letting you have your cake and eat it too, with a clear path to upgrade if you need to.
It’s a matter of reframing the whole experience. Rather than a chatbot that does everything at the same pace, Gemini is built to let you put in the kind of thought you want the model to put in before it gives you an answer.
User-controlled reasoning
There are two ways to do it: Standard and Extended. If you’re after a quick bit of info or a casual chat, Standard is what you want. But with Extended, the model will take its time to really look at your prompt before it replies.
According to Google, this is where you go for the hard stuff-multi-step problems or when you need a detailed breakdown. You might have to wait a little for the response, but you should be getting a more complete one. Of course, it all comes down to how you ask.
Availability and rollout
You can start using it on Gemini for the web, Android and the iPhone, whether you pay for it or not. Since it depends on the model, the controls you see may not be uniform right out of the gate.
“The new Thinking Levels are live on web, iOS and Android,” says Josh Woodward at Google, making it plain this isn’t a one-platform thing.
They have support for a number of models like the 3.1 Flash-Lite, 3.5 Flash and 3.1 Pro. In the demos we’ve seen, they’ve also been shown off with the 3.5 and 3.5 Flash-light variants.
A few things to keep in mind:
– Free and paid users can use both Standard and Extended
– What you see is up to the Gemini model
– The settings are in the model menu
– Works on the web, Android and iPhone
Deep Think and premium approach
For the most part, you’ll have your two modes, but there is a top-shelf option for some. Deep Think has been around since earlier in the year and is still just for the Google AI Ultra crowd on the 3.1 Pro.
Google puts it down to “maximum parallel reasoning.” Put simply, it throws more compute at the knottier, multi-layered issues. It’s a way of segmenting the product without leaving the free user with nothing.
Why now and what it signals
This is in the wake of I/O 2026 and the news on the 3.5 family. The 3.5 Flash is here, and 3.5 Pro is due in the next few weeks to unseat the 3.1 Pro as the flagship of the mainstream line.
Having these levels as the norm in Gemini is a shift. You can be fast when you have to, or switch to Extended for when you need to be sure. That kind of transparency can only help with confidence in the tool.
Give it a try by opening up the model settings in Gemini and selecting your mode. Just don’t be surprised if it takes a moment to chug through a complicated task in Extended mode.
It also means the style of the output is less about which model you picked and more about what you want. Google is taking the guesswork out of why you got a short or long answer and making the trade-off obvious.
What to watch next
Once 3.5 Pro is in the wild later this month, the lineup will be a different story. As these features make their way to more of the models, you’ll have to decide for yourself: is it time for a quick fix, or should you let Gemini put some thought into it?
