Nano Banana Pro is here: How to ditch Google’s old AI image generator for the new one

Google’s Nano Banana image generator has already gone viral because of how really, incredibly, and scarily good it is. But today, it’s old news already. Google just announced the much more impressive new Nano Banana Pro, based on the Gemini 3 Pro Image model, which is already putting good ol’ Banana to shame with its superpowers.

Nano Banana Pro is better with text generation, uses real-time data from Google, and applies advanced reasoning to create or edit images better than ever. It can also make spiffy new infographics, like this one I made below, which explains where exactly Nano Banana Pro is rolling out. But how do you turn on the new Gemini Nano Banana Pro model? Here’s the gist.

Gemini Nano Banana Pro is available globally already to everyone, free and paid subscribers alike, in the Gemini web and mobile apps. Free users will get limited quotas and then revert to Nano Banana, but paid subscribers of Google One’s AI Plus, AI Pro, and AI Ultra plans will have higher quotas to use the Pro model.

Google might show you a greeting when you open Gemini, giving you a quick way to turn on Nano Banana Pro, but if it doesn’t, you need to make sure you select two things in your Gemini app because Nano Banana Pro is only available with the Thinking model.

First, select 🍌 Create Image in the Gemini suggestion screen, as you usually do, then make sure to tap on the 2.5 Flash model button on the bottom right. Switch to the Thinking 3 Pro model. Once both of these options are selected correctly, Gemini will use the new Nano Banana Pro instead of the older Banana.

On the web, you need to do the same thing. Choose 🍌 Create Image and then click the model selector in the bottom right of the text entry box and choose Thinking with 3 Pro. This ensures Nano Banana Pro is activated and you can start asking Gemini to generate or edit the most complex images.

I’m still trying to wrap my head around how much more powerful Nano Banana Pro is. My colleague Mishaal has called it the AI Photoshop of his dreams in his hands-on, and I think even that is a bit of an understatement. I’m amazed to see it can understand slightly vague queries like what I gave it below, which generated the meme above, and can reason around complex text and generate compelling and accurate infographics like the Gemini availability table at the top of the post. And each time I asked it to make a minor edit, it fully understood what I meant and fixed that without messing with the rest.

I know what I’ll be using from now on to quickly visualize complex test results or ideas in my articles. It’s silly not to make use of such a tool when it’s so easy to use. Still, I’m terrified by how good it is and what other people will use it for. But hey, at least we have SynthID now to detect AI images.