1 hour ago
1 ARTICLE AD
Google is teasing better speed, lower battery use, and broader language support for on-device AI.
Google
Your next Android flagship may get a big Gemini Nano 4 boost, and Google’s already laying the groundwork. In a new developer preview, it’s pushing a faster, more efficient AI model that will power upcoming phones later this year.
The idea’s simple. Build apps now using the new Gemma 4 model, and that same code will carry over to supported devices when they arrive. It gives developers a head start while Google fine-tunes performance for real hardware.
This matters because AI’s becoming central to flagship phones. These models run locally instead of relying on the cloud, which means quicker responses and stronger privacy. Google’s clearly pushing this as a core layer for the next wave of Android devices.
There are still gaps. Google hasn’t named specific phones or pricing, and timing’s limited to later this year, but the direction’s clear.
Faster model, smaller footprint
Gemma 4 is the foundation for the next on-device model, with upgrades focused on efficiency. Google says it can run up to four times faster than earlier versions while using as much as 60 percent less battery.
Google
There are two variants, one for heavier reasoning and one for lower latency. The lighter version’s designed to feel more responsive, giving developers flexibility depending on their app needs.
It also supports more than 140 languages and handles text, images, and audio in one system, enabling features like translation and smarter assistants without needing a connection.
Why this matters for your phone
This marks a shift in how Android flagships will compete. Instead of leaning on cloud processing, Google’s pushing models that run directly on the device, tuned for newer AI chips from Qualcomm, MediaTek, and Google.
Gemini on a smartphone Unsplash
That should make features like assistants and transcription feel more reliable. It also gives phone makers room to differentiate based on hardware performance.
There’s a tradeoff. Performance will vary depending on whether a device supports AICore and newer accelerators. On unsupported hardware, the experience may fall back to slower CPU processing.
What to watch next
Google’s opening early access through its AICore preview, giving developers time to test apps before the rollout reaches consumer devices.
More features are planned during the preview, including better prompt controls and structured outputs.
Devices are expected in the coming months, though which phones will support it first is still unclear. If you’re planning to upgrade, the key question will be how well your device can actually run these models day to day.
Paulo Vargas is an English major turned reporter turned technical writer, with a career that has always circled back to…
Microsoft no longer wants to borrow its AI, it wants to build it
Frontier AI is the goal, and Microsoft is pulling out all the stops to get there by 2027.
Microsoft has been pushing AI on consumers whether they wanted it or not. Given the ferocity with which the company has been pushing AI into its products, you might be surprised to learn that it didn’t use its own AI. It took OpenAI’s technology, wrapped it into Copilot and Teams, and called it a day.
But things are changing. Whether the company noticed the public’s negative reaction to its bloated Windows 11 operating system or saw Linux gaining market share in gaming, Microsoft is finally working to introduce a calmer Windows 11 and focus on developing its own AI models.
This new AI attack steals models without touching the system
A side-channel attack can reconstruct AI models from a distance using leaked signals.
AI systems have long been treated like sealed black boxes, especially in areas like facial recognition and autonomous driving. New research suggests that protection isn’t as solid as assumed.
A KAIST-led team shows that AI systems can be reverse engineered remotely using emissions that leak during normal operation, without direct intrusion. Instead, the approach listens.
This wild MacBook Neo water-cooling mod turns it into a much faster machine
A liquid-cooled MacBook Neo sounds stupid until you see the performance gains
The MacBook Neo was never meant to be a powerful laptop for heavy workloads. It was built as a simple, affordable notebook that promises decent performance and solid battery life for everyday use. It is not supposed to need custom water cooling like a gaming PC.
And yet, that is exactly what happened.
