Products You May Like
After some leaks and preemptive announcements, Google has revealed the full Pixel 9 lineup, which is more extensive this year than ever before. Google has four new phones, all available for pre-order today. There are a number of important hardware differences, but one thing they all have in common is deep integration with Google Gemini.
Much like the Google I/O keynote, the presenters at the Pixel unveiling could barely open their mouths without saying “Gemini” or “AI.” Virtually everything new this year is tied to Google’s generative AI model. Large language models (LLMs) have yet to prove their worth, but Google’s new AI-first phones have the best shot yet of making the case.
A Pixel Quartet
The multi-tiered phone lineup is new for Google, which has previously released just one or two phones at a time. This time, we’ve got three flat phones and a new foldable. The Pixel 9 family starts with the $799 Pixel 9, which has a 6.3-inch 1080p screen. There’s a Pixel 9 Pro ($999) at the same size with a sharper screen and three cameras instead of two. Then you have the Pixel 9 Pro XL, which has a 6.8-inch OLED and the same enhanced camera setup as the 9 Pro. This device hits a new high price of $1,099.
Google’s second-generation foldable is the Pixel 9 Pro Fold. The external screen is almost identical to the Pixel 9, but inside is an 8-inch folding 2,076-by-2,152 OLED panel. Google says it’s the thinnest foldable in the world, but it’s also among the most expensive at $1,799.
All the new Pixel phones have Google’s Tensor G4 chip inside, which has substantially improved AI processing capabilities. The chip can output up to 45 tokens per second, twice as many as the Pixel 8’s Tensor G3. This relatively minor improvement turns out to be core to the Pixel 9 experience. Gemini on the Pixel 9 family will have more access to the data in your Google apps and what’s on your screen, but it doesn’t sacrifice your privacy to do it.
Gemini at Heart
Google increased the RAM on the new Pixel phones to 12GB for the base model and 16GB in all the Pro variants. You won’t get access to all of that memory, though. Google has decided to dedicate some RAM (it isn’t saying how much) to running the Gemini Nano model locally on the phone. This helps the traditional chatbot-style Gemini features, but it also improves features in the camera, keyboard, and several new experiences based on generative AI.
By running the Gemini Nano model on-device, you can parse personal data more securely. This has enabled Google to borrow a page from Microsoft’s playbook with the new Pixel Screenshots app. Every screenshot you take is now accessible from this app and scanned with Gemini to extract important details, making the content easily searchable. You can even ask questions about the content of your screenshots. Similarly, Gemini Nano uses the new Pixel Weather app’s data to generate (again, without going to the cloud) an instant summary of conditions right at the top of the app.
Google’s Call Assist features have long been a key selling point of the Pixel line, and they’re getting an AI-powered enhancement with the Pixel 9. Using the on-device model, you can have Gemini record and summarize phone calls, refer back to a previous conversation, and search for specific details.
Google hasn’t spent as much time talking about its image generation technology, a diffusion model called Imagen 3, but it’s front and center on the Pixel 9 series. A new Pixel Studio app lets you create (almost) anything you can imagine. Still, it’s not allowed to generate people yet because of Google’s well-documented issues in that area. You can also load in the photos you’ve taken and make extensive generative edits to them. However, this feature requires an internet connection.
Where’s the Killer App?
It should not come as a surprise that Google is investing heavily in AI for its new phones—building more powerful LLMs is the company’s primary activity lately. It’s not alone dumping money into generative AI, with the likes of Microsoft, OpenAI, and Facebook also building out their AI platforms. These companies are spending billions of dollars on the AI accelerators it takes to train and run these models with very little to show for it.
Credit: Google
But even after more than a year of AI hysteria, there’s no “killer app” for AI. There are some fairly impressive chatbots, and plenty of phones can rewrite your text messages to sound Shakespearian. But actually being helpful? That’s a bridge too far when every chatbot comes with a warning that it might make up details while being annoyingly confident in its wrongness.
Google’s control over the Pixel experience—from the core hardware, to the OS, to the apps that run on it—allows it to plug your data into the model. This feels like a better use for generative AI than making general web queries or asking for party planning advice. Instead, the model focuses on the data that is important to you. It’s hard to say if this will finally make all those billions well-spent, but Google’s at least got a shot.