Zero’s New Mini PCs Put a 160 TOPS AI Accelerator and 24GB Dedicated NPU Memory in a Tiny Desktop
Zero is taking a different approach to the fast-growing AI mini PC market. Instead of asking the main processor or integrated graphics to handle local AI tasks, the company’s latest compact desktops include a separate AI accelerator with its own dedicated memory.
The new Zero SEi14 AI and SEi13 AI have appeared on JD.com, and both are designed around a dual-system concept. One side of the machine works like a traditional x86 mini PC, running the operating system and everyday apps. The other side is a discrete NPU accelerator card built specifically for local AI inference.
That dedicated AI card is the headline feature. It is rated at 160 TOPS and comes with 24GB of LPDDR5 memory reserved only for AI workloads. This means model weights, quantized inference, text processing, multimodal tasks, and long-context calculations can run separately from the host CPU and integrated GPU.
For users interested in local AI models, this design could be a major advantage. Many AI PCs rely on shared system memory, which can limit performance when the operating system, applications, and AI workloads are all competing for the same resources. Zero’s new design gives the NPU its own memory pool, helping keep AI processing isolated and potentially more consistent.
The Zero SEi14 AI is the higher-end model. It uses an Intel Core Ultra 9 185H processor with 16 cores, 22 threads, boost speeds up to 5.1GHz, and 24MB of L3 cache. Graphics are handled by Intel Arc integrated graphics with 8 Xe cores running up to 2.35GHz. This model also includes Thunderbolt 4, giving it a stronger edge for users who need high-speed data transfer, display output, or power delivery support.
The Zero SEi13 AI is the more affordable option. It comes with an Intel Core i7-13620H processor featuring 10 cores, 16 threads, boost speeds up to 4.9GHz, and 24MB of L3 cache. Instead of Intel Arc graphics, it uses Intel UHD Graphics with 64 execution units at up to 1.5GHz. It also drops Thunderbolt 4 and replaces it with a standard 10Gbps USB-C port that supports data and video.
Despite the differences in processor and connectivity, both mini PCs use the same 160 TOPS NPU card and the same 24GB LPDDR5 dedicated AI memory. Both also support W4A8 and W8A8 quantized inference formats, which are commonly used to make local AI models more efficient by reducing memory and compute requirements.
The SEi14 AI comes with 32GB of DDR5 SO-DIMM memory, with two slots supporting speeds up to 5600 MT/s. The SEi13 AI includes 32GB of LPDDR5 memory in a 16GB x2 configuration. Both models include a 1TB M.2 2280 PCIe 4.0 x4 SSD, with support for storage expansion up to 4TB.
Connectivity is similar across the two systems. Both include WiFi 6 through an Intel AX200 module, Bluetooth 5.2, 2.5G Ethernet, two USB 2.0 ports, two USB 3.2 10Gbps ports, DisplayPort 1.4a, HDMI, and two 3.5mm audio jacks. The major difference is that the SEi14 AI adds Thunderbolt 4 at 40Gbps with DisplayPort 1.4 and power delivery support, while the SEi13 AI uses another 10Gbps USB-C connection.
Both systems are impressively compact, measuring 135 x 135 x 44.7 mm. They use a 19V/6.32A power adapter, making them small enough for a desk setup, home lab, AI experimentation station, or compact workstation environment.
Pricing places these machines firmly in the premium mini PC category. The Zero SEi14 AI with 32GB system memory, 1TB storage, and 24GB dedicated NPU memory is listed at CNY 13,988, which is approximately USD 2,067. The SEi13 AI with the same storage and NPU configuration is listed at CNY 11,895, or roughly USD 1,757.
Zero is also offering a five-year warranty and a 30-day trial period for both models. However, at the time of listing, both configurations are shown as out of stock.
What makes these mini PCs stand out is not just the raw 160 TOPS rating, but the way Zero has separated AI processing from the rest of the system. By giving the NPU its own 24GB memory pool and allowing it to handle model inference independently, the SEi14 AI and SEi13 AI are positioned as compact local AI machines rather than ordinary mini desktops with AI branding.
For developers, researchers, and enthusiasts who want to run local AI workloads without relying on cloud services, Zero’s new AI mini PCs could be an interesting glimpse at where compact desktop computing is headed.
