GMKtec EVO-X2 (Ryzen AI Max+ 395)
No editorial image yet — generic vendor mark shown. Credentials in spec table below.
The cheapest mainstream 128GB Strix Halo mini-PC (~$1,499 entry). Ryzen AI Max+ 395, 128GB LPDDR5X-8000 unified, Radeon 8060S iGPU, up to ~96GB allocatable as VRAM. One of the most-reviewed local-AI boxes of 2025-26 and often cited as the only sub-$1,500-class device that runs 70B with no dGPU.
Affiliate disclosure: as an Amazon Associate and partner of other retailers, we earn from qualifying purchases. The verdict on this page is our editorial opinion; affiliate links never influence what we recommend.
Sub-scores sum to 219 / 1000. Headline = 219 × 0.70 (Estimated-confidence discount) = 153. This is an algorithmic performance-tier score — distinct from, and often lower than, the editorial “Our verdict” below, which weighs value and real-world fit (especially for hardware we haven’t measured yet). How scoring works →
Extrapolated from 256 GB/s bandwidth — 25.6 tok/s estimated. No measured benchmarks yet.
Plain-English: Doesn't fit modern chat models usefully — vision models won't fit.
Verdicts extrapolated from catalog VRAM + bandwidth + ecosystem flags. Hover any chip for the rationale. Want measured numbers? Submit your own run with runlocalai-bench --submit.
What it does well
The EVO-X2 is the value entry to the 128GB unified-memory class. For roughly $1,499 (base config; 128GB variants run higher) you get the same Ryzen AI Max+ 395 silicon as pricier Strix Halo boxes, with up to ~96GB usable as VRAM — enough to run 70B models and even Qwen3-235B-class MoE locally with no discrete GPU. It's been extensively benchmarked (Tom's Hardware, ServeTheHome, Notebookcheck) and is the default 'cheapest way into big-model local inference' recommendation.
Where it struggles
Same Strix Halo bandwidth ceiling (~256 GB/s) means token speed is modest relative to capacity — great for fitting models, merely okay for fast generation. Networking is the EVO-X2's specific weak point: single 2.5GbE versus the dual-10GbE on rivals like the Beelink GTR9 Pro or Minisforum MS-S1 Max, which matters if you cluster or serve. ROCm/Vulkan-only, Linux-first, no CUDA — expect setup effort.
Bottom line
The best price/capability on-ramp to 70B+ local inference on x86. Pick it for the lowest cost of entry to 128GB unified; step up to Framework (repairable/open) or a dual-10GbE box (MS-S1 Max / GTR9 Pro) only if you need those specific angles.
Overview
The cheapest mainstream 128GB Strix Halo mini-PC (~$1,499 entry). Ryzen AI Max+ 395, 128GB LPDDR5X-8000 unified, Radeon 8060S iGPU, up to ~96GB allocatable as VRAM. One of the most-reviewed local-AI boxes of 2025-26 and often cited as the only sub-$1,500-class device that runs 70B with no dGPU.
Some links above are affiliate links. We may earn a commission at no extra cost to you. How we make money.
Specs
| System RAM (typical) | 128 GB |
| Power draw (peak) | 140 W |
| Released | 2025 |
| MSRP | $1499 |
| Backends | ROCm Vulkan |
Models that fit
Open-weight models small enough to run on GMKtec EVO-X2 (Ryzen AI Max+ 395) with usable context.
Frequently asked
Does GMKtec EVO-X2 (Ryzen AI Max+ 395) support CUDA?
Where next?
Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify hardware specifications.