Tiiny AI has introduced the AI Pocket Lab mini-PC, a very compact device approximately the size of a smartphone, albeit thicker, which has already secured a place in the Guinness World Records as The Smallest MiniPC (100B LLM Locally). This makes it the smallest computer capable of locally running a large language model with 100 billion parameters. To be precise, the company claims it manages 120 billion parameters.

Full details are not yet available, but it is known that the device is powered by a 12-core processor based on the Arm v9.2 architecture. Additionally, there appears to be a heterogeneous single-chip system and a discrete NPU (Neural Processing Unit), collectively delivering a performance of 190 TOPS (Tera Operations Per Second).

Considering the compact size, the inclusion of 80 GB of LPDDR5X memory is particularly remarkable. It also features a 1 TB SSD. The rise of cloud AI has brought about significant progress but has also created dependence, vulnerability, and resilience issues. With the Tiiny AI Pocket Lab, we believe intelligence should belong not to data centers, but to the people. This is the first step to making cutting-edge AI genuinely accessible, private, and personalized by bringing the capabilities of large models from the cloud to individual devices.

The system consumes about 65 watts, so the mini-PC comes with active cooling. The dimensions are 142 x 80 x 25.3 mm, with a weight of 300 grams. No information on the pricing is available yet.
Industry Reactions
Experts in the field are closely watching Tiiny AI’s new development due to its groundbreaking size and local processing capabilities. Such a device could revolutionize personal computing, offering more privacy and reducing reliance on remote servers.
Technical Comparisons
Comparatively, many current mini-PCs and AI-capable devices in the market often require larger setups to manage similar computational tasks, often relying on cloud integration. However, the Pocket Lab’s core strength lies in its ability to process significant AI workloads directly on a device small enough to hold in your hand.
Energy Consumption Insight
Regarding its energy use, drawing 65W places the Pocket Lab in a competitive range for mini-PCs, considering typical laptops with AI capabilities often draw similar power but without on-device AI processing to this extent.