192GB max RAM · 12 bays · 2 x 10GbE
AINAS Score 64
No verified price

AI ReadyQNAP
Why AI Ready: Xeon E-2378 with PCIe Gen 4 x8 expansion and a 128GB memory ceiling meets the Xeon ready rule; the best listed GPU is the RTX 3060 and QNAP documents no GPU inference on this model.
Updated
Can the QNAP TS-h3087XU-RP run local AI?
Partly. The QNAP TS-h3087XU-RP runs local models on Intel Xeon E-2378 alone, and its 128GB ceiling allows 70B class models at higher-quality quantization, or large mixture-of-experts models at 4-bit (Estimated), but replies will be slow without graphics acceleration.
A graphics card can be added through its PCIe x8 slot, which is the biggest single speed upgrade for local models.
Local LLM support is documented by the manufacturer.
The TS-h3087XU-RP suits a business or serious home lab that wants a large rackmount NAS which can also host local AI in containers. It has 30 bays (24x 3.5-inch plus 6x 2.5-inch), an 8-core Xeon E-2378 and 64GB of DDR4 ECC memory, expandable to 128GB. Docker and virtual machines are supported. A PCIe Gen 4 x8 slot gives a path to a graphics card. Memory is large enough that Mistral Large 2 123B fits at Q4 (Estimated). Two 10GbE and two 2.5GbE ports handle network traffic, and the warranty is 5 years. It scores 63 on the AINAS Score.
The main catch is that the TS-h3087XU-RP has no integrated graphics and no NPU, and QNAP documents no GPU inference on this model. Anyone who wants fast model responses without adding hardware is affected, because all model memory here is system RAM. Buyers in that position should look for a system with a supported GPU already fitted, or a slot and power budget that suit a card with enough VRAM for the models they want. The two 800W power supplies and slots that take dual-slot cards up to 300mm do leave room for the RTX 3060 Twin Edge OC on QNAP's list, though its VRAM is not stated.
Amazon US
TS-h3087XU-RP-E2378-64G
Ships from the US; delivery and import charges shown at checkout
$11,999.00
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 | QNAP Ollama guide applies (x86_64, 16GB or more for 7B models, GPU optional and must be in Container Station mode). Widest GPU list of this group, including RTX 3060, RTX 2060 6G and RTX 3050 6G with Container Station. RTX 3060 VRAM is not stated on the list. |
| Ollama | Yes | Vendor confirmed | QNAP Container Station Ollama tutorial for QTS and QuTS hero on x86_64. |
| Open WebUI | Yes | Vendor confirmed | QNAP Open WebUI tutorial, QTS 5.2 or QuTS hero h5.2 or later. |
| RAG | Limited | Unverified | Qsirch lists on-premises RAG with no model named. Not confirmed for this model. |
| AI search | Yes | Vendor confirmed | Qsirch AI Semantic Search: 64-bit x86, 5.0.1 or later, 8GB RAM. Qsirch says the AI Semantic Index and AI Mode need an x74 series NAS or later; not confirmed for this model, and AI Mode also needs a 16GB or larger VRAM GPU. |
| GPU expansion | Limited | Vendor confirmed | PCIe Gen 4 x8 slot, up to 300 mm dual-slot cards, 2x 800W PSUs. GPU list includes RTX 3060 Twin Edge OC, RTX 2060 6G, RTX 3050 6G, RTX A400 4GB. |
| Hardware acceleration | Limited | Vendor confirmed | No integrated graphics; spec page says graphics optional via a PCIe adapter. Coral Edge TPU on the AI accelerator list; listed GPUs support Hardware Transcoding. |
| AI photos | Yes | Vendor confirmed | QuMagie listed under QTS only. |
| AI video | Yes | Vendor confirmed | QVR Face listed under QTS only. |
| Surveillance AI | Yes | Vendor confirmed | QVR Surveillance 2 free channels plus 6 on registration; QVR Pro 8 free channels (QTS only). |
Calculated from the TS-h3087XU-RP'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 | Q8 | Yes | 19GB | Not measured | Estimated |
| Qwen3 8BQwen, 8.19B | Q8 | Yes | 19GB | Not measured | Estimated |
| Gemma 3 12BGemma, 11.8B | Q8 | Yes | 24GB | Not measured | Estimated |
| Qwen3 14BQwen, 14.8B | Q8 | Yes | 28GB | Not measured | Estimated |
| Phi-4 14BPhi, 14.7B | Q8 | Yes | 28GB | Not measured | Estimated |
| Mistral Small 3 24BMistral, 23.6B | Q8 | Yes | 38GB | Not measured | Estimated |
| Gemma 3 27BGemma, 27B | Q8 | Yes | 44GB | Not measured | Estimated |
| Qwen3 30B A3BQwen, 30.5B | Q8 | Yes | 47GB | Not measured | Estimated |
| Qwen3 32BQwen, 32.8B | Q8 | Yes | 50GB | Not measured | Estimated |
| DeepSeek-R1 32B (Qwen distill)DeepSeek, 32.8B | Q8 | Yes | 50GB | Not measured | Estimated |
| Llama 3.3 70BLlama, 70.6B | Q8 | Yes | 98GB | Not measured | Estimated |
| Llama 4 Scout 17Bx16ELlama, 109B | Q4 | Yes | 89GB | Not measured | Estimated |
| Qwen2.5 72BQwen, 72.7B | Q8 | Yes | 101GB | Not measured | Estimated |
| Mistral Large 2 123BMistral, 123B | Q5 | Yes | 112GB | Not measured | Estimated |
One verdict per workload, from the TS-h3087XU-RP's use-case scores and capability rows. Unknown means we cannot confirm it, not that it probably works.
The TS-h3087XU-RP is a good fit for local LLM work. The vendor confirms Ollama through Container Station, though local LLM support is marked limited, and Llama 3.3 70B fits at Q4 and Q8 (Estimated).
Private RAG is an OK fit on the TS-h3087XU-RP. The RAG capability comes from Qsirch, which names no model and is Unverified for this system, though Open WebUI is vendor confirmed.
AI search is a good fit for the TS-h3087XU-RP. Qsirch AI Semantic Search is vendor confirmed, but Qsirch says its AI Semantic Index and AI Mode need an x74 series NAS or later, and that is not confirmed here.
AI photos is a good fit for the TS-h3087XU-RP. QuMagie is vendor confirmed, but it is listed under QTS only, so choose QTS rather than QuTS hero for it.
Surveillance is a good fit for the TS-h3087XU-RP. QVR Surveillance gives 2 free channels plus 6 on registration, and QVR Pro gives 8 free channels, both under QTS only.
Home lab use is a good fit for the TS-h3087XU-RP. It supports Docker and virtual machines, takes up to 128GB of ECC memory, and offers three PCIe Gen 4 slots for add-in cards.
Business use is a good fit for the TS-h3087XU-RP. It has ECC memory, two 800W power supplies, 30 bays, 2 x 10GbE and a 5 year warranty.
Creative AI is a good fit for the TS-h3087XU-RP, with caveats. GPU expansion is limited but vendor confirmed: the PCIe Gen 4 x8 slot takes cards up to 300 mm, and the GPU list includes the RTX 3060 Twin Edge OC.
Out of the box the TS-h3087XU-RP has 64GB; the official ceiling is 128GB of DDR4 ECC UDIMM with ECC support. A memory upgrade is the cheapest way to make it more useful for AI.
Working backwards from 128GB: 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 200B (Estimated).
ECC support protects against silent memory errors on a box that runs around the clock.
It can. A 3x PCIe Gen 4 (2 x4, 1 x8); slot 1 cards up to 260 x 131.15 x 54.16 mm, slot 2 up to 300 x 131.15 x 54.16 mm, slot 3 up to 300 x 131.15 x 18.76 mm is the route, and a card with 16GB or more of its own memory changes what the TS-h3087XU-RP can run. Measure the space and check the 800W power budget before ordering.
For a personal library, yes. A compact embedding model and a small chat model will run, but the TS-h3087XU-RP is better as the document store behind a RAG system whose model runs on stronger hardware.
The manufacturer also provides AI search across your files (documented by the manufacturer), which covers simple document questions without setting up your own stack.
Built-in networking: 2 x 10GbE + 2 x 2.5GbE. Plenty for a shared knowledge base with several users.
| Configuration | RAM | SSD | Notes |
|---|---|---|---|
| TS-h3087XU-RP-E2378-64G | 64GB | None | Xeon E-2378 8C/16T, 64GB ECC. |
The closest QNAP model by bay count and AINAS Score. Only the specifications that differ are listed.
Its strongest AINAS profiles are Home Lab at 72 and RAG at 67, which points to home lab owners who want one box for storage, containers, media and a few AI services.
Other AI Ready systems closest in price.
Buying guides that list it
Workloads it suits
Yes, in memory terms. Mistral Large 2 123B needs 73GB at Q4 and fits within the 128GB ceiling (Estimated). At Q8 it needs 130GB and does not fit (Estimated). Qwen3 235B A22B does not fit at either level (Estimated). These are calculations, not benchmarks, and the system has no GPU fitted by default.
Yes. It ships with 64GB of DDR4 ECC UDIMM and the memory is expandable up to 128GB. The record marks the memory as expandable, so you are not limited to the amount fitted at purchase.
Yes. It uses DDR4 ECC UDIMM memory, and ECC is listed as supported. The Xeon E-2378 processor is paired with 64GB of ECC memory in the configuration sold, with a ceiling of 128GB.
Yes, from a listed set. The PCIe Gen 4 x8 slot takes cards up to 300 mm, dual-slot, and QNAP lists the RTX 3060 Twin Edge OC, RTX 2060 6G, RTX 3050 6G and RTX A400 4GB. Support is marked limited, and the system has no integrated graphics.
No. It has 0 built-in M.2 slots. NVMe storage is optional through a QM2 PCIe adapter, which would occupy one of the three PCIe Gen 4 slots (2 x4, 1 x8) in the system.
Yes. Container Station runs Docker on QTS 6.1.0 and QuTS hero h6.1.0, and virtual machines are supported too. QNAP's Ollama tutorial for x86_64 is vendor confirmed and asks for 16GB or more for 7B models, which the 64GB fitted covers with room to spare. A GPU is optional but must be in Container Station mode, and QNAP's list for this model includes the RTX 3060, RTX 2060 6G and RTX 3050 6G with Container Station. Open WebUI is vendor confirmed for QTS 5.2 or QuTS hero h5.2 or later.
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 2 x 10GbE + 2 x 2.5GbE built in, and the 3x PCIe Gen 4 (2 x4, 1 x8); slot 1 cards up to 260 x 131.15 x 54.16 mm, slot 2 up to 300 x 131.15 x 54.16 mm, slot 3 up to 300 x 131.15 x 18.76 mm can take a faster network card. The fastest port is 2 x 10GbE.
Every field records its source. Specifications were last checked on . Spot an error? Send a correction with a manufacturer link.