In a groundbreaking announcement, Tiiny AI Inc., a US-based deep-tech AI startup, has unveiled the Tiiny AI Pocket Lab, certified by Guinness World Records as the world's smallest personal AI supercomputer.
The device was officially verified in the category of "The Smallest MiniPC (100B LLM Locally)," marking a significant milestone in making advanced AI accessible to individuals without relying on traditional cloud infrastructure.
The unveiling event, held in Hong Kong, showcased the device's capability to run a full 120-billion-parameter Large Language Model (LLM) entirely on-device. Unlike traditional AI systems that depend on servers, high-end GPUs, or internet connectivity, the Pocket Lab operates independently, directly addressing key industry challenges: sustainability, privacy, and accessibility.
🔒 A Shift Toward Personal and Private AI
Tiiny AI's innovation comes as cloud-based AI faces increasing scrutiny over high energy consumption, rising costs, global outages, and privacy risks. The Pocket Lab offers an alternative, emphasizing personal, portable, and fully private intelligence.
Samar Bhoj, GTM Director of Tiiny AI, highlighted the company's vision:
"Cloud AI has brought remarkable progress, but it also created dependency, vulnerability, and sustainability challenges. With Tiiny AI Pocket Lab, we believe intelligence shouldn't belong to data centers, but to people. This is the first step toward making advanced AI truly accessible, private, and personal, by bringing the power of large models from the cloud to every individual device."
The device is engineered for energy efficiency, operating within a 65W power envelope, significantly lower than traditional GPU-based systems. Tiiny AI claims the Pocket Lab can handle over 80% of real-world AI needs in the "golden zone" of 10 billion to 100 billion parameters, with scalability up to 120 billion parameters—delivering intelligence comparable to models like GPT-4o.
🌟 Key Features and Specifications
The Tiiny AI Pocket Lab is tailored for a wide range of users, including developers, researchers, creators, professionals, and students. It supports multi-step reasoning, deep context understanding, agent workflows, content generation, and secure processing of sensitive data—all offline. It also offers long-term personal memory storage with bank-level encryption, ensuring user data remains private and persistent.
Feature | Details |
|---|---|
Processor | ARMv9.2 12-core CPU |
| AI Compute Power | Custom heterogeneous module (System-on-Chip + Dedicated Neural Processing Unit), delivering approx. 190 TOPS |
Memory & Storage | 80 GB LPDDR5X + 1 TB Solid State Drive |
Model Capacity | Runs up to 120B-parameter Large Language Models fully on-device |
Power Efficiency | 30 W Thermal Design Power, 65 W typical system power |
| Dimensions & Weight | 14.2 × 8 × 2.53 cm, approx. 300g (pocket-sized) |
Ecosystem | One-click deployment of dozens of open-source Large Language Models and agent frameworks |
Connectivity | Works fully offline; no internet or cloud required |
The Pocket Lab's performance is driven by two proprietary technologies: TurboSparse, a neuron-level sparse activation technique that enhances inference efficiency, and PowerInfer, an open-source heterogeneous inference engine that dynamically distributes computations across the CPU and NPU, achieving server-grade performance at minimal power levels.
🌐 Ecosystem and Company Background
Tiiny AI has built a robust open-source ecosystem around the Pocket Lab, allowing one-click installation of leading models such as OpenAI GPT-OSS, Llama, Qwen, DeepSeek, Mistral, and Phi. It also supports seamless deployment of popular AI agents. Users can expect continuous updates, including official over-the-air (OTA) hardware upgrades, with additional features slated for release at CES in January 2026.
Founded in 2024, Tiiny AI Inc. comprises a team of global engineers from prestigious institutions and industry leaders. Their mission is to pioneer personal AI supercomputing, bringing cloud-grade intelligence to on-device, private, and offline environments.
This development could reshape the AI landscape, shifting power from centralized data centers to individual users.
 |  |  |
 |  |  |