The American company Tiiny AI, specialized in advanced technology, has launched the new “Tiiny AI Pocket Lab” device, which has officially earned the title of “the world’s smallest personal AI supercomputer” according to the Guinness World Records.
The device is characterized by its small power-board size and its ability to run massive Large Language Models (LLMs) featuring up to 120 billion parameters locally, without the need for a cloud connection or the use of traditional GPUs.
Through this innovation, the company seeks to empower individuals to benefit from advanced AI tools without relying on massive data centers, while delivering performance equivalent to that of supercomputers, reducing energy consumption, and ensuring high privacy.
The device was launched on December 10th as a practical solution to sustainability challenges and high energy costs associated with traditional AI infrastructure.
Samar Bhoj, GTM Director at Tiiny AI, explained in a press release: “Despite the great progress made by cloud AI, it faces challenges related to reliability, security, and sustainability. With Tiiny AI Pocket Lab, we believe that AI should be available to individual users, becoming more private and suitable for every device.”
Multiple Capabilities and Uses
The device is designed to meet various personal AI use needs, targeting creators, developers, researchers, and students. It provides advanced features such as multi-step reasoning, context understanding, smart agent workflow management, content creation, and secure processing of sensitive data without the need for the internet. It also stores data locally using bank-level encryption, which enhances privacy and provides long-term memory.
The device supports running models ranging from 10 to 100 billion parameters and can run models up to 120 billion parameters, providing performance close to the GPT-4 model in multi-step analysis while maintaining data security.
Advanced Technical Specifications
The device relies on a 12-core ARMv9.2 processor with 65 watts of power, providing high performance with energy efficiency and a low carbon footprint compared to traditional systems. It utilizes two main technologies:
-
TurboSparse: Activates only the required neurons to improve efficiency without affecting performance.
-
PowerInfer: An open-source engine that distributes AI workloads between the Central Processing Unit (CPU) and the Neural Processing Unit (NPU) to reduce energy consumption.
The device operates on an open-source system, supporting the installation of popular models such as Llama, DeepSeek, and GPT-OSS, and provides users with regular updates and upgrades online. These features are scheduled to be showcased at the Consumer Electronics Show (CES) in January 2026.
United News Network – UNN Arabic
An independent media platform providing reliable news and objective analysis, seeking to promote peace and cultural dialogue around the world to convey the truth and build bridges of understanding between peoples.
For more news, you can visit our homepage:
