Minisforum AI X1 Pro-470

What it is

Minisforum’s AI X1 Pro-470. Ryzen AI 9 HX 470 (Gorgon Point: 4 Zen 5 + 8 Zen 5c, 12C/24T, up to 5.2 GHz), Radeon 890M (16 RDNA 3.5 CUs, up to 3.1 GHz), XDNA 2 NPU (55 TOPS, ~86 total). Dual-channel DDR5-5600 SODIMM (up to 128 GB), three M.2 PCIe 4.0 slots (two x4, one x1; up to 12 TB), dual 2.5GbE, dual USB4, HDMI 2.1, DP 2.0, OCuLink (PCIe 4.0 x4 for eGPU), Wi-Fi 7, fingerprint, built-in 135 W PSU. 195 × 195 × 47.5 mm, ~1.5 kg. Ships with Windows 11 Pro on configured SKUs. Same silicon class as the Beelink SER10 MAX and other HX 470 minis.

Who it is for

People who want an upgradeable Gorgon Point mini with OCuLink, triple M.2, and dual 2.5GbE for local 7B–14B models, light agents, or a compact creator / lab box, and who will populate dual-channel RAM. Wrong if you need Strix Halo capacity (24B–70B class at usable speed), CUDA-native, or the highest memory bandwidth. The dual SODIMMs and OCuLink are the points if you plan to add RAM or an eGPU later; the bandwidth ceiling is not.

What it actually runs

• Comfortable: 7B–14B dense quantized at usable speed on CPU or Vulkan iGPU. Smaller MoE and agent workloads fit easily.
• Possible: 20B–35B class quantized on 64–128 GB system RAM if you accept that ~90 GB/s is the wall. Decode stays memory-bound. Same limits as other HX 470 boxes.
• Not the win: dense 70B at practical speed, training, or matching a Strix Halo / DGX Spark box on capacity. Ignore “86 TOPS” and vendor LLM marketing slides. The NPU is still mainly a Windows Copilot / limited-path feature for local GGUF stacks.

Standout details

• Price snapshot mid Sep 2026: Minisforum US and Amazon list barebone around $759 (was higher list). Configured 32 GB + 1 TB tracks ~$1,300–$1,360; 32 GB + 2 TB around $1,399. Higher-RAM SKUs appear on some channels but DDR5 pricing keeps them elevated. Treat as a snapshot.
• Memory is dual SODIMM DDR5-5600. Theoretical dual-channel bandwidth is ~89–90 GB/s. That is roughly one-third of the LPDDR5X-8000 bus on Strix Halo minis. Official support reaches 128 GB; practical dual-channel configs matter more than the paper max.
• OCuLink is the real expansion differentiator versus sealed cubes and some other HX 470 minis. Triple M.2 and user-replaceable RAM are the other practical upsides.
• Same silicon class as the Beelink SER10 MAX. Incremental clock and NPU gains over the prior HX 370; not a generational leap in local LLM bandwidth.
• Cooling claims quiet sustained operation with PCM + heat pipes. Real loads still move air; the form factor is larger than the smallest cubes but still desk-friendly.

Honest verdict

A solid, upgradeable Gorgon Point mini with OCuLink and triple M.2 for people who want 7B–14B local models or a compact expandable box and will buy dual-channel RAM. It is not a Strix Halo machine. Bandwidth is the limiter. If you need 24B–70B capacity at desk-friendly speed, look at the Max+ 395 cubes (Framework, EVO-X2, Bosgame, GTR9 Pro, MS-S1 MAX) or a CUDA brick. If you just need a quiet, networked, openable mini with eGPU path, this is fine—same class as the SER10 MAX.