惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Apple Machine Learning Research
Apple Machine Learning Research
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
G
Google Developers Blog
博客园 - 司徒正美
J
Java Code Geeks
aimingoo的专栏
aimingoo的专栏
A
About on SuperTechFans
博客园 - 三生石上(FineUI控件)
WordPress大学
WordPress大学
T
The Blog of Author Tim Ferriss
D
Docker
大猫的无限游戏
大猫的无限游戏
D
DataBreaches.Net
腾讯CDC
V
Visual Studio Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
C
Check Point Blog
M
MIT News - Artificial intelligence
Jina AI
Jina AI
I
InfoQ
雷峰网
雷峰网
The Cloudflare Blog
美团技术团队
Engineering at Meta
Engineering at Meta

Android Authority

I know YouTube Music is flawed, yet I prefer it over Spotify Survey reveals 50% of users don’t like the new Google Health app It’s time for Samsung’s S Pen to evolve or die The Motorola Moto G Stylus (2026) is a sequel we didn’t need NotebookLM is quickly becoming the podcast app I didn’t know I needed Samsung’s next Galaxy Watch update could finally make your health data useful Google’s Gemini Spark is ready to run your digital errands while your phone is off Telegram’s finally getting an official Wear OS app again Nintendo is back on mobile, and it wants to turn your selfies into minigames Google Drive’s big document scanner overhaul is finally here — don’t overlook its power Spotify will finally give you real profile tools to make music listening more social Acer’s new gaming handheld might dodge the worst of tech inflation Meta is cooking up a new line of smart glasses, and they may not be Ray-Bans ChatGPT is retiring this beloved legacy model in June Is Microsoft Copilot not working? Here’s what’s going on (Update: Back up) Samsung Gallery starts quietly ending OneDrive support ahead of schedule Here’s a first look at custom wallpapers in Google Messages Rivian is pretty sure customers want AI, not Android Auto Leaked iPhone 18 Pro dummy units may have just shown the next Android phone color trend A company spent $500 million in one month after forgetting to set AI usage limits Now even MediaTek’s cheap chips are embarrassing the Tensor G5 in one major area Pixel 10 Pro XL user says Google returned their phone worse than dead The best robot pool cleaners of 2026: Top picks for all budgets and pool sizes Claude Opus 4.8 is more honest, less deceptive, and considerably cheaper Roborock’s Qrevo Curv 2 Flow is ready to mop up the competition — and your filthy floors Google is making it easier to share Gemini chats, media, and more with your team One UI 9 borrows one of the iPhone’s most useful call features This is the biggest mistake Oura is making with the Oura Ring 5 This Verizon user owed $400, but the carrier made an unexpected move Google’s Fitbit Air makes a strong case for minimalism and ditching your smartwatch
CrankGPT is a private AI chatbot that's powered entirely ...
Chethan Rao · 2026-06-19 · via Android Authority
CrankGPT screenshot YouTube

TL;DR

  • A Europe-based company has built a tiny and fully private AI chatbot known as CrankGPT.
  • Unlike conventional AI chatbots, CrankGPT is powered entirely by a hand crank.
  • The company is currently accepting requests to demonstrate the gadget, though you can’t buy one yet.

Whether it’s ChatGPT, Google Gemini, or some other big-tech chatbot, these tools have become an inseparable part of modern life. As user demand for chatbots grows, so does the need for more natural resources. A Europe-based company, Squeez Labs, has now revealed a chatbot concept that aims to make generative AI cost-effective and less resource-intensive, using a simple hand crank.

Cleverly named CrankGPT, this little box has no battery and relies entirely on a crank to generate power. To put it simply, it behaves like a conventional chatbot but is powered entirely by turning the crank. Inside this small box is a 20W hand-crank generator powered by a stock Raspberry Pi 5 with 8GB of RAM (via Gizmodo).

This efficient arrangement allows CrankGPT to recognize what you say, process it, and then deliver the answer using the Piper voice model. As one would expect, small models are best suited to run on the CrankGPT, with the creators recommending the 350M and 1.2B variants of Liquid AI LFM2, or Gemma (1B).

As the video above shows, the CrankGPT is anything but quiet. However, this is a small compromise for a device that is supposed to be both private and energy-efficient. While the creators have reportedly tried to generate images, poetry, and even code using CrankGPT, the device is evidently better suited for small, everyday queries.

The developers have also integrated a cooling fan to keep any heat in check. They add that the crank can get “a lot harder to turn” when speech synthesis and LLM inference happen simultaneously.

Perhaps the best argument for a device of this size and capability is that it runs on a fully local model that is always under the user’s control.

“It offended our European small-practical-car sensibilities to see people around us throwing kilowatts and thousands of tokens at tasks small models could accomplish just as well as huge ones, for a fraction of the cost and energy,” the CrankGPT creators said.

The CrankGPT website lays out the case for this product pretty well, though there’s no information related to a broader launch or a potential price tag. You can request a demo of the device, however.

Regardless of whether this product becomes more than a concept one day, it shows that there are ways to use AI without it impacting our natural resources. However, it’s worth remembering that training these AI models also bears a heavy burden on the earth’s resources, so until that part’s figured out, these solutions can only go so far.

But with that being said, the CrankGPT could be the start of something bigger, especially as AI becomes a more integral part of our lives, for better or worse.

Thank you for being part of our community. Read our Comment Policy before posting.