Minisforum announced two Panther Lake-based mini PCs, the M2 Pro and MS-03, aimed at local AI workloads (Gizmochina; Notebookcheck; Videocardz). According to Minisforum's product blog, the M2 Pro can be configured with up to an Core Ultra X9 388H and delivers up to 180 TOPS of combined CPU+NPU+GPU compute (Minisforum blog; Videocardz). The MS-03 is a workstation-class chassis with Panther Lake CPUs and an NPU rated up to 50 TOPS (Notebookcheck). Reported hardware highlights include an internal power supply, 10G and 2.5G LAN, WiFi 7, OCuLink for eGPU, a Microsoft Copilot button and microphone array, and support for up to 128 GB LPDDR5X or similar memory configurations across models (Gizmochina; Notebookcheck). Minisforum and trade coverage positioned the systems for "local-first" or hybrid AI workflows, citing OpenClaw compatibility on the M2 Pro (Minisforum blog; Videocardz).
What happened
Minisforum introduced two new Intel Panther Lake mini PCs, the M2 Pro and the MS-03, alongside announcements for NAS models in the same product family (Notebookcheck; Gizmochina). The company showcased the Panther Lake systems at Intel Core Ultra Series 3 promotional events and product briefings earlier this year (Videocardz; Minisforum blog). Notebookcheck and Gizmochina report that the M2 Pro targets compact local-AI workloads while the MS-03 targets workstation-style use cases and small server roles (Notebookcheck; Gizmochina).
Technical details
According to Minisforum's blog and product coverage, the M2 Pro can be configured with an Core Ultra X9 388H and offers up to 180 TOPS of aggregate compute across CPU, NPU5, and integrated Xe3 GPU (Minisforum blog; Videocardz; Notebookcheck). The product is reported to support up to 128 GB of LPDDR5X-class memory, three M.2 PCIe 4.0 slots, an OCuLink port for eGPU expansion, a front-panel Microsoft Copilot button and microphone array, and an internal power supply instead of an external brick (Minisforum blog; Gizmochina; Notebookcheck).
Technical details (I/O and configuration)
Coverage lists multiple high-speed networking and I/O options: one 10G and one 2.5G LAN jack, WiFi 7, three USB-A and three USB4 ports, HDMI and DisplayPort outputs, an SD card slot, and VESA-mount capability for the metal-chassis M2 Pro (Gizmochina; Notebookcheck). The MS-03 uses a different Panther Lake SKU such as a Core Ultra 7 356H, raises the platform thermal envelope to higher TDPs for workstation duties, and provides an NPU around 50 TOPS (Videocardz; Notebookcheck).
Industry context
Industry context: Coverage frames these releases as part of a broader movement toward "local-first" AI hardware, where vendors combine on-device NPUs, beefed-up integrated GPUs, and high-throughput I/O to run models offline or in hybrid mode (Minisforum blog; Videocardz; Notebookcheck). Observed patterns in similar product launches emphasize thermal design, memory bandwidth, and storage density as key constraints when manufacturers advertise multi-hundred-TOPS totals across heterogeneous compute.
For practitioners
For practitioners: the emphasis on OCuLink eGPU support, multiple M.2 slots, and 10G networking indicates these boxes are designed for workflows that may combine local model inference, fast dataset access, and occasional cloud offload. Industry reporting suggests the M2 Pro is marketed to run OpenClaw in a hybrid configuration, moving sensitive context and reasoning on-device while using cloud models selectively (Minisforum blog; Videocardz).
What to watch
What to watch: real-world thermals and sustained throughput when all compute engines are active; software support and optimized runtimes for mixing Intel NPU, Xe GPU, and CPU work; and final retail SKUs, pricing, and availability for the M2 Pro and MS-03, since coverage currently mixes press blog claims with early show demos (Minisforum blog; Gizmochina; Notebookcheck).
Scoring Rationale #
This is a notable product announcement for practitioners building local or hybrid AI setups: the 180 TOPS claim and workstation I/O make these mini PCs useful testbeds, but they are not a paradigm-shifting platform release. The story matters to hardware-focused ML engineers and edge deployment teams.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.