Tiiny AI Inc. (www.tiiny.ai), a US-based AI startup, has launched the Tiiny AI Pocket Lab. Guinness World Records has verified it as the smallest MiniPC capable of running a 100B-parameter LLM locally.
This is the first time a pocket-sized device can run a 120-billion-parameter large language model entirely on-device. It does not need cloud access, servers, or high-end GPUs.
Tiiny AI Pocket Lab uses up to 65W of power. It can run large models while using less energy than traditional GPU-based systems.
Cloud-based AI has issues with energy consumption, outages, high costs, and privacy concerns. Tiiny AI Pocket Lab is a portable device that runs models locally and keeps data private. According to Tiiny AI, the main problem now is not computing power but dependence on the cloud.
“Cloud AI has brought remarkable progress, but it also created dependency, vulnerability, and sustainability challenges,” said Samar Bhoj, GTM Director of Tiiny AI. “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.”
Tiiny AI Pocket Lab can be used by developers, researchers, creators, professionals, and students. It supports multi-step reasoning, agent workflows, content generation, and the secure processing of sensitive data without an internet connection. User data, preferences, and documents are stored locally with strong encryption.
Crucially, Tiiny AI Pocket Lab operates in the ‘golden zonTiiny AI Pocket Lab runs models in the 10B to 100B parameter range, which covers most real-world needs. It can also run models up to 120B parameters, similar to GPT-4o, and does all processing offline on the device.
According to Grand View Research, the global LLM market size was estimated at USD 7.4 billion in 2025 and is projected to reach USD 35.4 billion by 2030, growing at a CAGR of 36.9% from 2025 to 2030. It is driven by advances in self-supervised learning, domain-specific model development, and rising enterprise adoption across sectors such as retail, e-commerce, healthcare, and media.
Tiiny AI Pocket Lab is powered by two core technology breakthroughs that enable large-parameter models to run on a compact device. TurboSparse, a neuron-level sparse activation technique, significantly improves inference efficiency while maintaining full model intelligence. PowerInfer, an open-source heterogeneous inference engine with more than 8,000 GitHub stars, accelerates heavy LLM workloads dynamically distributing computation across CPU and NPU, enabling server-grade performance at a fraction of traditional power consumption. Together, these technologies allow Tiiny AI Pocket Lab to deliver capabilities that previously required professional GPUs costing thousands of dollars.
Key Specifications:
| Processor | ARMv9.2 12-core CPU |
| AI Compute Power | Custom heterogeneous module (SoC + dNPU), delivering ≈190 TOPS |
| Memory & Storage | 80GB LPDDR5X + 1TB SSD |
| Model Capacity | Runs up to 120B-parameter LLMs fully on-device |
| Power Efficiency | 30W TDP, 65W typical system power |
| Dimensions & Weight | 14.2 × 8 × 2.53 cm, approx. 300g, pocket-sized |
| Ecosystem | One-click deployment of dozens of open-source LLMs and agent frameworks |
| Connectivity | Works fully offline; no internet or cloud required |
Tiiny AI Pocket Lab supports one-click installation of open-source models like OpenAI GPT-OSS, Llama, Qwen, DeepSeek, Mistral, and Phi. It also works with open-source AI agents such as OpenManus, ComfyUI, Flowise, Presenton, Libra, Bella, and SillyTavern. Users get regular updates, including hardware upgrades. These features will be available at CES in January 2026.
The team was formed in 2024 and includes engineers from MIT, Stanford, HKUST, SJTU, Intel, and Meta. Their research has appeared in top academic conferences. In 2025, Tiiny AI raised a multi-million dollar seed round.