16GB max RAM · 4 bays · 2 x 2.5GbE
AINAS Score 32
£636
Amazon UK · Verified

AI CapableASUSTOR
Why AI Capable: Intel Celeron N5105 (4C/4T) with Intel UHD Graphics, no NPU, no PCIe or GPU expansion and a 16GB memory ceiling; runs ASUSTOR's Ollama app for small CPU-only models at best.
Updated
Can the ASUSTOR Flashstor 12 Pro (FS6712X) run local AI?
Partly. The ASUSTOR Flashstor 12 Pro (FS6712X) runs local models on Intel Celeron N5105 alone, and its 16GB ceiling allows 7B to 8B class models (Estimated), but replies will be slow without graphics acceleration.
No graphics card can be fitted, which keeps every model on the CPU.
Local LLM support is documented by the manufacturer.
The Flashstor 12 Pro suits buyers who want a compact all-flash NAS with 12 NVMe slots and a 10GbE port, and who treat local AI as a small extra. It has no drive bays, an Intel Celeron N5105 with four cores, and 4GB of DDR4-2933 SO-DIMM memory that can be raised to 16GB. It runs ADM 5 with Docker and virtual machines. ASUSTOR's Ollama + Open WebUI app is vendor confirmed, but CPU-only. Qwen3 8B fits at Q4 (Estimated). The AINAS Score is 36, which reflects a storage-first machine rather than an AI one. It also carries a 3 year warranty, and photo recognition and surveillance are both vendor confirmed.
The limit that matters is the 16GB memory ceiling combined with no GPU path. Buyers who want 14B models with headroom, or anything larger, will hit it. Qwen3 14B is only marginal at Q4 (Estimated), and Qwen3 32B does not fit at Q4 or Q8 (Estimated). There is no PCIe slot, no OCuLink, no Thunderbolt or USB4 and no NPU, so the ceiling cannot be worked around. Those buyers should look for a higher memory maximum, a PCIe slot or other external GPU connection, or an NPU with a documented role in inference.
Amazon UK
4GB
£634.99
Verified
In stock
Buy on AmazonProduct 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 | Vendor confirmed | ASUSTOR App Central lists Ollama + Open WebUI (v0.11.4.r01, ADM 5.0+) for the FS67 series. CPU-only inference; 4GB base RAM and 16GB max limit model size. The app does not install a model by default to avoid failures on some NAS models. |
| Ollama | Yes | Vendor confirmed | Ollama + Open WebUI app lists FS67 as a supported series; installs via Docker. |
| Open WebUI | Yes | Vendor confirmed | Bundled with the Ollama + Open WebUI App Central package. |
| RAG | Limited | Vendor confirmed | ASUSTOR markets NAS as RAG knowledge base storage with vector DBs (Chroma, pgvector, Weaviate) via Docker; Open WebUI includes RAG. ASUSTOR recommends external compute for inference. |
| AI search | No | Unverified | No AI or semantic file search feature found in ADM 5.1 material. |
| GPU expansion | No | Vendor confirmed | No PCIe expansion slot on spec page and no iGPU. ASUSTOR states its NAS models do not feature a discrete GPU. |
| Hardware acceleration | Limited | Vendor confirmed | Intel UHD Graphics iGPU (Intel ARK) used for 4K H.265 hardware decoding/transcoding per ASUSTOR; no NPU. GPU use for AI inference not documented. |
| AI photos | Yes | Vendor confirmed | Photo Gallery 4.2 (ADM 5.1.1+) lists FS67 series; ADM 5.1 adds facial recognition grouping powered by locally run AI. |
| AI video | No | Unverified | No AI video feature found in ADM documentation. |
| Surveillance AI | Yes | Vendor confirmed | Surveillance Center 2.8.4 (app 1663) and Surveillance Center 3 (app 1661) list FS67; up to 44 IP cameras per ASUSTOR spec page licence table. |
Calculated from the Flashstor 12 Pro (FS6712X)'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 | Q8 | Yes | 13GB | Not measured | Estimated |
| Llama 3.1 8BLlama, 8.03B | Q5 | Yes | 15GB | Not measured | Estimated |
| Qwen3 8BQwen, 8.19B | Q4 | Yes | 15GB | Not measured | Estimated |
| Gemma 3 12BGemma, 11.8B | Q4 | Marginal | 15GB | Not measured | Estimated |
| Qwen3 14BQwen, 14.8B | Q4 | Marginal | 16GB | Not measured | Estimated |
| Phi-4 14BPhi, 14.7B | Q4 | Marginal | 16GB | 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 Flashstor 12 Pro (FS6712X)'s use-case scores and capability rows. Unknown means we cannot confirm it, not that it probably works.
The Flashstor 12 Pro is a poor fit for running local language models. Support is limited: the vendor-confirmed Ollama app is CPU-only, and the 4GB base memory and 16GB maximum limit the size of model you can load.
Private RAG is a poor fit. ASUSTOR rates RAG as limited here, with vector stores such as Chroma possible through Docker, and the vendor itself recommends external compute for inference rather than this CPU.
AI search is a poor fit, and we cannot confirm it works. No AI or semantic file search feature was found in the ADM 5.1 material, and that finding is marked Unverified.
AI photo handling is an acceptable fit. Photo Gallery 4.2 lists the FS67 series, and ADM 5.1 adds facial recognition grouping powered by locally run AI, vendor confirmed.
Surveillance is an acceptable fit. Surveillance Center 2.8.4 and Surveillance Center 3 both list the FS67 series, and the vendor spec page licence table allows up to 44 IP cameras.
Home lab use is acceptable. Docker is supported, VirtualBox lists this series for virtual machines, and the 12 NVMe slots and 10GbE port suit fast shared storage, though four cores and 16GB cap how much you can run.
Business use is acceptable for fast flash file storage. It has 12 NVMe slots, a 10GbE port and a 3 year warranty, but the memory is not ECC and there are no drive bays.
Creative AI is a poor fit. There is no PCIe slot and no discrete GPU, and the Intel UHD Graphics iGPU is documented for 4K H.265 transcoding, not for AI inference.
Out of the box the Flashstor 12 Pro (FS6712X) has 4GB; the official ceiling is 16GB of DDR4-2933 SO-DIMM. A memory upgrade is the cheapest way to make it more useful for AI.
Working backwards from 16GB: reserve about 8GB for the operating system and storage services, and at roughly 0.6GB per billion parameters at 4-bit, the largest model that fits is around 13B (Estimated).
There is no supported way to add a GPU to the Flashstor 12 Pro (FS6712X): no PCIe slot and no OCuLink. Inference runs on the CPU, with Intel UHD Graphics handling media tasks, so keep to small models.
For a personal library, yes. A compact embedding model and a small chat model will run, but the Flashstor 12 Pro (FS6712X) is better as the document store behind a RAG system whose model runs on stronger hardware.
Built-in networking: 10GbE. Plenty for a shared knowledge base with several users.
| Configuration | RAM | SSD | Notes |
|---|---|---|---|
| 4GB | 4GB | None | Diskless; 1x 4GB DDR4 pre-installed, 2 SO-DIMM slots. |
The closest ASUSTOR model by bay count and AINAS Score. Only the specifications that differ are listed.
Its strongest AINAS profiles are Home Lab at 44 and Business at 43, which points to home lab owners who want one box for storage, containers, media and a few AI services.
Look elsewhere if you are anyone who wants to run models larger than about 8B parameters, or buyers who expect to add a graphics card later.
Other AI Capable systems closest in price.
Buying guides that list it
Yes, as an estimate. Qwen3 8B needs 5.2GB at Q4 and fits (Estimated). At Q8 it needs 8.9GB and is marginal (Estimated), so it loads with little headroom. Gemma 3 12B is marginal at Q4 (Estimated) and does not fit at Q8 (Estimated). These are calculations, not benchmarks. Inference runs on the CPU only through ASUSTOR's Ollama + Open WebUI app, which is vendor confirmed.
Yes, up to 16GB. The 4GB configuration ships with one 4GB DDR4 SO-DIMM installed and has two SO-DIMM slots. The memory is DDR4-2933 and the maximum is listed as 16GB, made up of two 8GB modules. The memory is not ECC. A larger memory pool raises the model sizes that fit (Estimated), but 16GB is the hard ceiling.
No. There is no PCIe expansion slot, no OCuLink port, no Thunderbolt and no USB4. ASUSTOR states its NAS models do not feature a discrete GPU. The only graphics hardware is the Intel UHD Graphics iGPU, which is documented for 4K H.265 hardware decoding and transcoding. Its use for AI inference is not documented.
12. The Flashstor 12 Pro has 12 NVMe slots and no conventional drive bays, so all storage is flash. It pairs those slots with one 10GbE port. It does not list any PCIe expansion slot. The 4GB configuration is sold diskless, so you supply the NVMe drives yourself.
Yes. Docker is supported on this series under ADM 5, and ASUSTOR's Ollama + Open WebUI app installs via Docker. Virtual machines are also listed: VirtualBox supports the FS67 series on ADM 5.1.3 and later. With four CPU cores and a 16GB memory maximum, containers and virtual machines share limited resources.
Only marginally, as an estimate. Qwen3 14B needs 9.3GB at Q4, which is marginal within the 16GB ceiling (Estimated), so it loads with very little room left for the NAS itself. At Q8 it needs 16GB and does not fit (Estimated). Qwen3 32B does not fit at either level (Estimated). These are calculations, not benchmarks, and the 4GB configuration needs a memory upgrade to two 8GB modules first. Inference runs on the CPU only.
Yes. Ollama support is documented by the manufacturer. See the capability table above for the source.
Yes. Open WebUI support is documented by the manufacturer. See the capability table above for the source.
Yes. AI photos support is documented by the manufacturer. See the capability table above for the source.
Yes. Surveillance AI support is documented by the manufacturer. 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 10GbE built in. The fastest port is 10GbE.
Every field records its source. Specifications were last checked on . Spot an error? Send a correction with a manufacturer link.