Asus unveiled the NUC 16 Pro—a compact PC powered by Intel's Panther Lake processors, designed for local AI model work.

Asus Unveils the NUC 16 Pro Mini-PC

The company Asus has released a new compact computer – the Mini‑PC NUC 16 Pro. The device is powered by Intel Panther Lake processors, including the flagship Core Ultra X9 388H with Arc B390 graphics and an integrated neural processing unit (NPU) delivering 50 TOPS.

What Makes It Stand Out
* Edge Computing and Local LLM Deployment

Asus positions the NUC 16 Pro as a platform for working with large language models and edge‑processing tasks. The combined CPU + GPU + NPU performance in machine‑learning mode reaches 180 TOPS.

* Compact Size

The chassis volume is only 0.7 L. Dimensions: 144 × 122 × 41 mm. The case is designed so that two M.2 slots can be easily accessed without tools, simplifying upgrades.

* Memory and Energy Efficiency

Built‑in LPDDR5x up to 96 GB provides a +20 % performance boost and 50 % lower power consumption compared with the previous NUC generation.

Ports and Connectivity
Front panel | Rear panel
2 × USB‑A 3.2 Gen2 | 2 × Thunderbolt 4
1 × USB‑C Gen2 | 1 × HDMI 2.1 (or DisplayPort)
2 × USB‑A 3.2 | 2 × LAN 2.5 Gbps | Wi‑Fi 7, Bluetooth 6.0

The dual‑fan cooling system keeps the unit stable even under heavy load.

Software
* Asus AI SuperBuild – an intuitive interface for running large language models locally without internet connectivity. This enhances data privacy and allows sensitive information (medical records, financial data) to be processed around the clock.

* Built‑in fTPM 2.0 protection adds an extra layer of security.

Configurations on Sale
In China, NUC 16 Pro variants with a 16‑core Core Ultra X7 358H processor are already available (similar core count to the X9 388H but at lower frequencies). This configuration comes with 32 GB RAM and a 1 TB SSD, priced at about $1,600.

Thus, the Asus NUC 16 Pro combines powerful hardware, compactness, and thoughtful edge‑AI support, making it an attractive choice for professionals who need local processing of large language models with high performance and security.