16GB max RAM · 4 bays · 10GbE
AINAS Score 28
£630
Amazon UK · Verified

AI CapableGMKtec
Why AI Capable: Intel N150 (no NPU), 12GB soldered LPDDR5, no GPU expansion path; Docker capable at best.
Updated
Can the GMKtec NucBox G9 run local AI?
The GMKtec NucBox G9 can run small local models, with limits. Its Intel N150 and 12GB memory ceiling suit small models up to around 7B parameters (Estimated), so treat it as a storage-first NAS that can host light AI tools.
There is no way to add a discrete GPU, so all inference runs on the CPU and basic integrated graphics.
Local LLM support is confirmed by community reports.
The NucBox G9 suits a buyer who wants a compact all-flash box for file storage and light container work, with small local models as a side interest. It has four M.2 slots, two 2.5GbE ports and an Intel N150 with 4 cores. Both network ports use Intel i226 controllers. Docker and virtual machines are both supported, and it ships with Windows 11 Pro and Ubuntu. Ollama is community confirmed on Windows. Llama 3.2 3B fits at Q4 (Estimated), and it is marginal at Q8 (Estimated). The AINAS Score is 28, which fits a system built for storage and light services rather than heavy AI work. It has no drive bays, so all storage is M.2 flash, and it carries a 1 year warranty. It comes with 12GB of memory and 512GB of SSD storage in its single configuration.
The limit that rules it out for some buyers is memory. The NucBox G9 has 12GB of LPDDR5 4800 soldered to the board, and it cannot be upgraded. Anyone who wants to run 8B class models comfortably will hit that ceiling: Llama 3.1 8B is marginal at Q4 (Estimated) and does not fit at Q8 (Estimated). Gemma 3 12B does not fit at Q4 (Estimated). There is also no way to add a GPU, because it has no PCIe slot, OCuLink or USB4. That affects anyone planning private RAG, larger models or creative AI work. Look instead for a system with upgradeable memory and a high memory ceiling, plus a GPU path such as a PCIe slot, OCuLink or USB4, or an NPU. Check the maximum memory figure, not just the amount included.
No verified price
We have not verified a listing for the GMKtec NucBox G9 at a retailer we track. The specifications below stand on their own; a price appears here as soon as a listing is confirmed.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon.co.uk or Amazon.com, as applicable, at the time of purchase will apply to the purchase of this product.
Each capability shows how it was confirmed: Verified by AINAS, Vendor confirmed, Community confirmed or Unverified.
| Capability | Supported | Evidence | Note |
|---|---|---|---|
| Local LLM | Limited | Community confirmed | No vendor AI software. Ships with Windows 11; Ollama has a native Windows build. 12GB soldered RAM limits it to small models. |
| Ollama | Yes | Community confirmed | Ollama offers a native Windows installer; Linux NAS OS installs are user choice. |
| Open WebUI | Yes | Unverified | Runnable via Docker or Python on Windows/Linux; no platform specific source found. |
| RAG | Limited | Unverified | Only via user installed tools such as Open WebUI; no vendor feature. |
| AI search | Limited | Unverified | No vendor NAS OS or AI search; third party apps only. |
| GPU expansion | No | Vendor confirmed | No PCIe slot, OCuLink or USB4. |
| Hardware acceleration | Limited | Vendor confirmed | Intel UHD Graphics 24 EU listed; no NPU. |
| AI photos | Limited | Unverified | No vendor photo app; third party apps such as Immich via Docker not verified on this hardware. |
| AI video | Limited | Unverified | No vendor feature; third party only, not verified. |
| Surveillance AI | Limited | Unverified | No vendor surveillance app; third party software only, not verified. |
Calculated from the NucBox G9's memory and each model's published size at each quantization. Rows marked Estimated are calculations, not benchmarks. Check any model.
| Model | Best quantization | Runs? | Memory needed | Expected performance | Verification |
|---|---|---|---|---|---|
| Llama 3.2 3BLlama, 3.21B | Q5 | Yes | 11GB | Not measured | Estimated |
| Llama 3.1 8BLlama, 8.03B | Q2 | Yes | 12GB | Not measured | Estimated |
| Qwen3 8BQwen, 8.19B | Q4 | Marginal | 12GB | Not measured | Estimated |
| Gemma 3 12BGemma, 11.8B | None fits | No | 18GB at Q4 | Not measured | Estimated |
| Qwen3 14BQwen, 14.8B | None fits | No | 20GB at Q4 | Not measured | Estimated |
| Phi-4 14BPhi, 14.7B | None fits | No | 19GB at Q4 | Not measured | Estimated |
| Mistral Small 3 24BMistral, 23.6B | None fits | No | 25GB at Q4 | Not measured | Estimated |
| Gemma 3 27BGemma, 27B | None fits | No | 29GB at Q4 | Not measured | Estimated |
| Qwen3 30B A3BQwen, 30.5B | None fits | No | 31GB at Q4 | Not measured | Estimated |
| Qwen3 32BQwen, 32.8B | None fits | No | 32GB at Q4 | Not measured | Estimated |
| DeepSeek-R1 32B (Qwen distill)DeepSeek, 32.8B | None fits | No | 32GB at Q4 | Not measured | Estimated |
| Llama 3.3 70BLlama, 70.6B | None fits | No | 60GB at Q4 | Not measured | Estimated |
| Llama 4 Scout 17Bx16ELlama, 109B | None fits | No | 89GB at Q4 | Not measured | Estimated |
| Qwen2.5 72BQwen, 72.7B | None fits | No | 65GB at Q4 | Not measured | Estimated |
| Mistral Large 2 123BMistral, 123B | None fits | No | 96GB at Q4 | Not measured | Estimated |
One verdict per workload, from the NucBox G9's use-case scores and capability rows. Unknown means we cannot confirm it, not that it probably works.
Local LLM work is a poor fit on the NucBox G9. Support is limited and community confirmed, and the 12GB of soldered memory restricts it to small models, with Llama 3.2 3B fitting at Q4 (Estimated).
Private RAG is a poor fit here. RAG is limited and unverified, working only through user installed tools such as Open WebUI, with no vendor feature, and Open WebUI itself is unverified, so nothing is confirmed on this hardware.
We cannot confirm AI search on the NucBox G9. The capability is marked limited and unverified, with no vendor NAS OS or AI search, so only third party apps would provide it, and none is confirmed.
AI photos is unknown for the NucBox G9, so we cannot confirm it. There is no vendor photo app, and third party apps such as Immich via Docker are not verified on this hardware, although Docker itself is supported.
Surveillance is unknown on the NucBox G9, and we cannot confirm it. The capability is limited and unverified, with no vendor surveillance app, and only third party software would do the job, none of it verified.
Home lab use is a poor fit for the NucBox G9. Docker and virtual machines are supported, but the Intel N150 has 4 cores and 12GB of soldered memory that cannot grow, which caps how many services you can run.
Business use is a poor fit for the NucBox G9. It has no ECC memory, no drive bays and no PCIe slot, only a 1 year warranty, and 12GB of soldered memory that cannot be expanded.
Creative AI is a poor fit for the NucBox G9. GPU expansion is vendor confirmed as not supported, with no PCIe slot, OCuLink or USB4, and hardware acceleration is limited to the Intel UHD Graphics with 24 EU and no NPU.
The NucBox G9 ships with 12GB of LPDDR5 4800 (soldered) that cannot be upgraded, so the configuration you buy is the one you keep. For local AI that matters more than anything else on the spec sheet: model size is capped by memory, not storage.
As a rule of thumb, a 4-bit model needs about 0.6GB per billion parameters plus room for context. After leaving around 8GB for the NAS operating system, 12GB leaves space for a model of roughly 6B parameters at most (Estimated).
No. The NucBox G9 has no PCIe slot or OCuLink port for a graphics card. Inference runs on the CPU, with Intel UHD Graphics (24 EU) handling media tasks, so keep to small models.
With limits. The NucBox G9 can host a small RAG stack such as AnythingLLM or Open WebUI with a compact embedding model, but memory and compute restrict how large the answering model can be. It suits a personal document library more than a team knowledge base.
Networking is 2 x 2.5GbE. Fine for one or two users; indexing a large archive over the network will be slower.
| Configuration | RAM | SSD | Notes |
|---|---|---|---|
| 12GB + 64GB eMMC + 512GB SSD | 12GB | 512GB |
The NucBox G9 scores best for Business (30/100) and Home Lab (28/100). It suits small offices that need shared storage with private AI tools and room to grow.
It is a weaker fit for anyone who wants to run models larger than about 8B parameters and for buyers who expect to add a graphics card later.
Other AI Capable systems closest in overall score.
Yes, at Q4. Llama 3.2 3B needs 2GB at Q4 and fits (Estimated), and at Q8 it needs 3.4GB and is marginal (Estimated), meaning it loads with little headroom. These are calculations, not benchmarks. It is the largest of the models listed on this page that fits at Q4 (Estimated). The NucBox G9 has 12GB of soldered memory. Ollama is community confirmed on Windows, while local LLM support overall is limited because the memory restricts it to small models.
No. The 12GB of LPDDR5 4800 is soldered to the board, so it cannot be expanded, and the maximum memory is the same 12GB that it ships with. Only one configuration is sold, with 12GB of memory, 64GB eMMC and 512GB of SSD storage. Buy it only if 12GB is enough for your use, because memory cannot be changed later. Storage is the part you can change, through the four M.2 slots.
No. GPU expansion is vendor confirmed as not supported. The system has no PCIe slot, no OCuLink port and no USB4 or Thunderbolt ports, so there is no route for an external or internal graphics card. Its only graphics hardware is the Intel UHD Graphics with 24 EU, and there is no NPU. Hardware acceleration is rated limited for that reason. Buyers who need a GPU should look for a system with a PCIe slot or OCuLink.
Yes. Docker is supported, and so are virtual machines. The NucBox G9 ships with Windows 11 Pro and Ubuntu. The 12GB of soldered memory and the 4 core Intel N150 are shared between the host, every container and every virtual machine, so keep workloads light. Third party AI apps run through Docker, such as Immich, are unverified on this hardware. Ubuntu is the Linux option.
Yes, two: Windows 11 Pro and Ubuntu. GMKtec supplies both, so you can use Windows for the native Ollama build, which is community confirmed, or Ubuntu as the Linux option. There is no vendor AI software, so file sharing, containers and any AI tools are your own setup. Docker and virtual machines are supported on the platform. Open WebUI, RAG, AI search, AI photos and surveillance are all Unverified on this hardware.
Yes, four. The NucBox G9 has four M.2 2280 slots on PCIe 3.0, each up to 4TB, though the lane count is not stated. There are no SATA drive bays, so M.2 flash is the only way to add storage. The single configuration comes with a 512GB SSD and 64GB of eMMC storage. Memory is separate and cannot be upgraded, so the slots add storage only.
Yes. Ollama support is confirmed by community reports. See the capability table above for the source.
Yes. Open WebUI support is not yet verified. See the capability table above for the source.
Partly. AI photos support is not yet verified. See the capability table above for the source.
Partly. Surveillance AI support is not yet verified. See the capability table above for the source.
Partly. Hardware acceleration support is documented by the manufacturer. See the capability table above for the source.
It has 2 x 2.5GbE built in. The fastest port is 2 x 2.5GbE.
Every field records its source. Specifications were last checked on . Spot an error? Send a correction with a manufacturer link.