“A Supercomputer in Your Pocket”: Tiiny AI Pocket Lab Enters Guinness and Promises Local LLMs up to 120B
The US startup Tiiny AI has unveiled the Tiiny AI Pocket Lab — a power-bank-sized device that Guinness World Records has officially recognized as a record holder in the category of “the smallest mini-PC capable of running LLMs 100B+ locally.” The product announcement and launch date are set for December 10, 2025.
According to the company, the Pocket Lab can run models with up to 120 billion parameters without relying on the cloud — a task that typically requires server-grade infrastructure. The value proposition is straightforward: privacy, autonomy, and full control over data (especially for enterprise use cases and field work), combined with a sustainability argument — reduced dependence on data centers and their growing energy appetite.
The specifications most frequently cited by the media read almost like those of a “mini rack”: dimensions of approximately 14.2 × 8 × 2.53 cm, weight around 300 g, up to 80 GB of LPDDR5X memory, a 1 TB storage drive, a 12-core ARM processor, and an integrated AI module rated at roughly 190 TOPS, with a stated power envelope of about 65 W. It is precisely this combination of memory capacity and NPU performance that is supposed to enable local operation of very large models through quantization and optimization.
At the same time, it is worth keeping a cool head: some marketing claims (such as references to “OTA hardware upgrades”) have already raised questions among commentators, and the bold promise of “120B locally” is so far based mainly on press releases and secondary reporting — without publicly available independent benchmarks. Still, the broader trend has been loudly marked: “intelligence from the cloud into the pocket” is no longer just a metaphor and is beginning to crystallize into a product category.
The company’s official website: tiiny.ai.
