MINISFORUM AI X1 Pro-370

The MINISFORUM AI X1 Pro-370 is a Copilot+ AI mini PC powered by the AMD Ryzen AI 9 HX 370, a Zen 5 processor with 12 cores and 24 threads, a 2.0 GHz base frequency, and a 5.1 GHz maximum frequency. Its XDNA 2 NPU delivers up to 50 TOPS, while total system AI performance reaches up to 80 TOPS. The 50 TOPS NPU exceeds the 40 TOPS Microsoft requires for Copilot+ certification, which allows AI writing assistance, summarization, translation, and other supported AI tasks to run on-device rather than through a cloud service. The system supports up to 128GB of DDR5 memory, three M.2 PCIe 4.0 NVMe slots, and an OCuLink port for external GPU expansion, with a built-in 135W power supply in place of an external brick. The chassis measures 195 × 195 × 47.5 mm and weighs 1.5 kg, and the system ships with Windows 11. It is positioned as the upgradeable alternative to soldered-memory AI systems, trading peak memory bandwidth for the ability to add capacity later as model sizes grow.The AI X1 Pro-370 supports up to 128GB of DDR5 memory across two SO-DIMM slots running at 5600 MT/s. Because the modules are replaceable rather than soldered, capacity can be increased after purchase, which matters for local AI work where the size of the model you can load is bounded by installed memory rather than by compute. Storage consists of three M.2 2280 NVMe slots, each supporting drives up to 4TB: two run at PCIe 4.0 x4 with sequential read speeds of roughly 6700 to 7300 MB/s, and the third runs at PCIe 4.0 x1 at roughly 1500 to 1900 MB/s, which suits it to archive or backup duty rather than active model loading. The three-slot layout separates an operating system drive, a working dataset drive, and a bulk library drive inside a single chassis, without external enclosures and without giving up a slot to the boot device, which is a common constraint on two-slot mini PCs.Graphics are handled by the integrated AMD Radeon 890M running at 2900 MHz, with flexible VRAM allocation drawn from system memory rather than a fixed pool. That allocation model matters for local AI work, because diffusion models and mid-sized language models can be given far more graphics memory than a fixed-VRAM discrete card at a similar price point would offer, and the ceiling rises as system memory is upgraded. The system drives up to four simultaneous 4K displays through one HDMI 2.1 FRL port and one DisplayPort 2.0 port, both supporting 8K at 60Hz or 4K at 120Hz, plus two USB4 ports at 4K 60Hz each. For workloads the integrated GPU cannot cover, the OCuLink port provides a direct PCIe 4.0 x4 connection to an external GPU dock, delivering higher bandwidth and lower latency than a USB4 or Thunderbolt eGPU enclosure, which makes it the preferred route for CUDA workloads or discrete-card gaming.Connectivity on the AI X1 Pro-370 includes two 2.5 Gigabit RJ45 Ethernet ports, Wi-Fi 7 with Bluetooth 5.4, two USB4 ports at 40Gbps, two USB 3.2 Gen2 Type-A ports, and one USB 2.0 Type-A port. The rear USB4 port supports 65W to 100W Power Delivery input as well as 15W output; the front USB4 port provides 15W output. The front panel also carries two USB 3.2 Gen2 Type-A ports, a 3.5mm combo jack, dual digital microphones, and a dedicated Copilot button, with a fingerprint verification button on the top surface for Windows Hello sign-in and an SD card slot on the side for direct media import. A Kensington lock slot and a clear-CMOS reset hole are located at the rear alongside the OCuLink port and the AC power connector. Power comes from an internal 135W supply rated at 19V/7.1A, so a single cable runs to the desk with no external power brick to find space for.Common use cases for the MINISFORUM AI X1 Pro-370 include running local large language models through Ollama or LM Studio, Stable Diffusion image generation, Copilot+ on-device features such as summarization and live captions, Proxmox and VMware virtualization hosts, home lab and self-hosted service nodes, software development environments needing many cores and a large memory pool, video editing with separate project, cache, and library drives, and multi-monitor office or trading setups running four 4K displays without a dock. Buyers typically choose the AI X1 Pro-370 when they want on-device AI acceleration, three NVMe slots, and upgradeable memory in a 1.5 kg chassis with an internal power supply. Buyers whose workloads need maximum memory bandwidth for very large models are better served by a Strix Halo system such as the MS-S1 MAX, buyers wanting higher AI performance should consider the AI X1 Pro-470 with a 55 TOPS NPU and up to 86 TOPS of total system AI performance, and buyers needing 10GbE networking should look at the MS-01, MS-03, or MS-A2.