Does your NAS actually fit AI?
A five-question diagnostic. Answer honestly about the box you have (or are about to buy), and the tool tells you which AI workloads are a no, a stretch, or a go — and when to add RAM, a GPU, or a separate box. As of 2026-11.
The diagnostic
RAM you "actually have" is the number after any upgrades — not the stock one. If you've already added memory, select that.
What each answer means
| RAM you have | Runs comfortably | Stretch / tight | Needs upgrade |
|---|---|---|---|
| 4 GB | Embedding / RAG, 1-3B | — | 7-8B, 14B+ |
| 8 GB | Embedding, 3B, 7-8B (tight) | 14B | 32B, 70B |
| 16 GB | 7-8B, 14B | 32B | 70B |
| 32 GB | 14B, 32B, MoE (GPT-OSS) | 70B (slow) | — |
| 64 GB+ | 32B, 70B | — | — |
A GPU (iGPU or dGPU) moves every "stretch" row left — the same model runs faster. CPU-only is always a no, just slower. Model-by-model sizes →
FAQ
Can my NAS run AI at all?
If it has a modern x86 or ARM CPU, almost certainly yes. The deciding factor is RAM: 4 GB runs small models, 8 GB handles a 7-8B model at a stretch, 16 GB is comfortable for 8B, 32 GB opens 14-32B, and 64 GB+ is where 70B-class models become real. A GPU or iGPU is a speed boost, not a requirement.
Do I need an NPU for AI on my NAS?
Only for specific vision workloads (camera person/vehicle detection via Frigate, or the NAS vendor\u2019s own AI photo feature). For running a chat or coding model (a local LLM), the NPU in current NAS hardware is mostly wasted — RAM and, if you want speed, a GPU matter more.
What if my NAS has too little memory?
Three options in order: (1) upgrade the RAM if the box has slots; (2) add an M.2 SSD to offload; (3) run the AI on a separate low-power mini-PC or GPU box and keep the NAS for storage. The last is the most flexible and cheapest to run.
Which AI is worth running on a NAS at all?
The highest-value, lowest-risk workloads are RAG over your own files (embedding models — tiny, fast, private) and light 7-8B chat assistants. Heavier 32-70B models are realistic only on memory-rich or DIY boxes. The diagnostic below ranks your box against each.