Hardware
GMKtec EVO-X2 (Ryzen AI Max+ 395) 128GB for local LLMs
128 GB system memory. Usable GPU memory in this catalog is AMD VGM dedicated memory, 96 GB.
Memory
128 GB unifiedCatalog fit uses 96 GB dedicated graphics memory from the row note, not the full system pool.
Bandwidth
256 GB/swww.amd.com/en/blogs/2025/faqs-amd-variable-graphics-memory-vram-ai-model-sizes-quantization-mcp-more.htmlPrice
$4,254€3,800
new (idealo, marketplace, delivery unknown, incl. shipping)
price as of 17 Oct 2026
idealo.deidealo data 17 Oct 2026 22:53. 128GB/2TB at €3,799.96 incl. shipping from GMKtec Direct, a marketplace offer, delivery unknown. The next regular-retailer offer was €3,979. AMD's VGM FAQ says up to 96 GB dedicated graphics memory, and that the iGPU can technically address 112 GB total on a 128 GB system. This catalog scores the 96 GB dedicated figure, not 112 GB and not the full 128 GB. Bandwidth 256 GB/s is from that FAQ. Unified memory is not the same as discrete VRAM.
Parts in the catalog
- GMKtec EVO-X2, Ryzen AI Max+ 395, 128 GB, 2 TB, as listed on idealo
Which models fit
Estimate at 8,192 tokens, or the model's maximum when that is shorter. full means the formula fits in usable memory. offload means GPU plus system RAM. no means it does not fit. This is not a measured file.
| Model | Q4Q4_K_Mestimate | Q5Q5 (~0.71 bytes/param)estimate | Q8Q8estimate | FP16FP16estimate |
|---|---|---|---|---|
| Llama 3.2 3B | full | full | full | full |
| Llama 3.1 8B | full | full | full | full |
| Mistral 7B v0.3 | full | full | full | full |
| Gemma 2 9B | full | full | full | full |
| Qwen2.5 14B | full | full | full | full |
| Phi-4 14B | full | full | full | full |
| Gemma 2 27B | full | full | full | full |
| Qwen2.5 32B | full | full | full | full |
| Llama 3.3 70B | full | full | full | no |
| Qwen2.5 72B | full | full | full | no |
| DeepSeek-R1-Distill 70B | full | full | full | no |
| NVIDIA Nemotron 3.5 Bolt 74B-A9B | full | full | full | no |
| Reflection Beam 1.0 | no | no | no | no |
| NVIDIA Nemotron 3.5 Super VL | full | full | no | no |
| DeepSeek V4.1 Flash 1017 | no | no | no | no |
| Meta Muse Spark Open | no | no | no | no |
We'll match you with vetted providers
Embed this
Badges cover nine scored GPUs. This machine is not one of them.
Common questions
- Can GMKtec EVO-X2 (Ryzen AI Max+ 395) 128GB run Llama 3.3 70B?
- Yes, fully in memory (estimate). Llama 3.3 70B at Q4_K_M needs about 46.0 GB at 8,192 tokens, and GMKtec EVO-X2 (Ryzen AI Max+ 395) 128GB has about 96.0 GB usable.
- Which models fit fully on GMKtec EVO-X2 (Ryzen AI Max+ 395) 128GB at Q4?
- 13 of 16 models fit fully at Q4 (estimate, 8,192 tokens): Llama 3.2 3B, Llama 3.1 8B, Mistral 7B v0.3, Gemma 2 9B, Qwen2.5 14B, Phi-4 14B, Gemma 2 27B, Qwen2.5 32B, Llama 3.3 70B, Qwen2.5 72B, DeepSeek-R1-Distill 70B, NVIDIA Nemotron 3.5 Bolt 74B-A9B, NVIDIA Nemotron 3.5 Super VL.
- How fast is GMKtec EVO-X2 (Ryzen AI Max+ 395) 128GB (estimate)?
- About 35 tok/s for Llama 3.2 3B at Q4 (estimate, not a benchmark).