Steep on-device requirements will limit Gemini Intelligence to premium phones

Google said on May 12 that Gemini Intelligence will begin rolling out this summer on select Samsung and Google phones, adding on-device AI features such as multi-step app automation, web summarization, smarter form filling, Gboard’s “Rambler,” and natural-language widget creation. The company described the package as coming first to its “most advanced devices,” but did not publish a broad compatibility list in its main announcement. (blog.google) Google’s security and product posts said the system is designed to process some tasks locally through Android’s AI stack while giving users controls over app access and permissions. (9to5google.com) Separate reporting on May 15 and May 16 said the hardware bar is high enough that many recent Android phones may not qualify. ### Which phones is Google actually promising first? Google’s public blog post named “select Samsung and Google phones” as the first devices to receive Gemini Intelligence this summer. The same post said broader availability across Android devices — including watches, cars, glasses and laptops — will come later this year. Google also said it had spent months fine-tuning multi-step automation on the Galaxy S26 and Pixel 10. Mindy Brooks, Google’s vice president of product management, wrote that Gemini Intelligence is meant to bring “the best of Gemini” to the company’s most advanced devices. Dave Kleidermacher, Google’s vice president for platforms security and privacy, said users will be able to turn app automation on for specific apps later this year and that purchases require user confirmation. (blog.google) ### What are the minimum specs that are causing concern? Google’s Gemini Intelligence page includes requirements that outside publications said are tucked into footnotes. Those requirements, as reported by 9to5Google and Android Authority, include a flagship chip, at least 12GB of RAM, support for AI Core, and support for Gemini Nano v3 or higher. (blog.google) The reports also said devices must be eligible for at least five Android OS upgrades, at least six years of quarterly security updates, and meet device-quality thresholds such as crash-rate standards. Android Developers documentation separately says Gemini Nano runs in Android’s AICore system service and is used for on-device generative AI tasks such as summarization, rewriting, image description and speech recognition. (blog.google) Google’s developer materials say on-device execution keeps prompts local and that inference speed depends on device hardware. ### Why are Pixel 9 and some Samsung phones showing up as possible exclusions? 9to5Google reported on May 15 that the Gemini Nano v3 requirement appears to exclude Google’s Pixel 9 series and Samsung’s Galaxy Z Fold 7, which the outlet said are still on Gemini Nano v2. Android Authority reported the same day that Pixel 7 Pro, Pixel 8 Pro, Pixel 9 series, Galaxy Z Fold 7 and Samsung’s TriFold appear set to miss out under the current criteria. (9to5google.com) Those conclusions were based on Google’s published requirements and a developer-page list of devices said to support Nano v3. Google has not, in the blog posts reviewed, issued a detailed public statement naming Pixel 9 or Galaxy Z Fold 7 as excluded models. (developer.android.com) The current gap is between Google’s broad launch language and third-party reporting that maps those requirements onto specific devices. ### What exactly will Gemini Intelligence do on the phone? Google said Gemini Intelligence will automate multi-step tasks across apps, summarize web content in Chrome, simplify complex form filling, rewrite spoken thoughts into polished messages through Rambler, and generate custom widgets from natural-language prompts. The company also described proactive features such as Magic Cue, which can surface suggestions based on context, though Google said users control whether data is shared and which apps Gemini can access. (9to5google.com) Google’s security post said the system uses Private Compute Core, Private AI Compute and protected KVM for some proactive assistance features. The company said AICore does not have direct internet access and that model downloads are routed through Private Compute Services. (blog.google) ### Where will the next hard details come from? Google’s next major Android event is Google I/O, which begins May 19, 2026, and the company has already said Gemini Intelligence features will roll out in waves starting this summer. The clearest near-term markers are whether Google publishes an official compatibility list, whether device makers confirm Nano v3 support, and which Samsung and Pixel models are named when the rollout starts. (blog.google) (mashable.com) (blog.google)