64GB max RAM · 10 bays · 2 x 10GbE
AINAS Score 43
£1,699
Amazon UK · Verified
AI CapableTerraMaster
Why AI Capable: Intel Xeon E-2224G (4 cores, UHD P630 iGPU, no NPU) with a 64GB memory ceiling. The PCIe 3.0 slot count and width are not confirmed, so the x8 GPU expansion the AI Ready line needs under 128GB is not shown.
Updated
Can the TerraMaster U24-722-2224 run local AI?
Only lightly. Expect 32B class models comfortably, and 70B class models at 4-bit with little headroom (Estimated) on the TerraMaster U24-722-2224: Intel Xeon E-2224G does the work with up to 64GB of memory, which is fine for experiments rather than a daily assistant.
Without any GPU option, the processor carries the whole inference load.
Local LLM support is not yet verified.
The U24-722-2224 suits a buyer who wants a 24-bay rackmount NAS for bulk storage and a modest amount of local AI on the side. It has an Intel Xeon E-2224G with 4 cores, and memory sits in slots rather than being soldered: it takes DDR4-3200 non-ECC UDIMM up to 64GB, from 8GB included. TOS has a Docker manager, so Ollama and Open WebUI can run as containers, though TerraMaster does not name this model for either. Llama 3.3 70B fits at Q4 (Estimated), and models up to Qwen3 32B fit at both Q4 and Q8 (Estimated). The AINAS Score is 43. Treat it as a storage server first, with CPU-only AI as an extra, since no GPU path is confirmed.
The main thing that rules it out is the lack of a confirmed GPU path. No GPU expansion is listed. The PCIe 3.0 expansion slots are described for 10GbE and RAID cards, the slot count is not confirmed, and GPU use is not documented. Anyone planning image generation, heavy RAG indexing or models that need more than a 4-core CPU can offer should look for a system with confirmed x8 PCIe expansion for a graphics card, or an NPU. Buyers moving large files to 24 bays should also note that networking is four 1GbE ports. Look for built-in 10GbE or faster if transfer speed matters to you.
No verified price
We have not verified a listing for the TerraMaster U24-722-2224 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 | Unverified | CPU-only inference on a 4-core Xeon E-2224G via Docker. TerraMaster blog covers Ollama on other x86 models; no source for this model. |
| Ollama | Limited | Vendor confirmed | TerraMaster blog describes running Ollama as a container through the TOS Docker manager on x86 NAS generally; this model not named. |
| Open WebUI | Limited | Vendor confirmed | Same blog names Open WebUI as a container for x86 TOS NAS; this model not named. |
| RAG | Limited | Unverified | Possible through Open WebUI or similar containers; no source showing RAG on this model. |
| AI search | No | Unverified | No local AI or semantic file search documented for this model. |
| GPU expansion | Limited | Unverified | PCIe 3.0 slots described for 10GbE and RAID cards; GPU use not documented. |
| Hardware acceleration | Limited | Unverified | CPU has UHD Graphics P630 per Intel; TerraMaster use on this model not confirmed. No NPU. |
| AI photos | Limited | Unverified | TOS 7 Photos: AI face recognition and 80+ scene categories. Requires TOS 7; no source confirms TOS 7 for this model. |
| AI video | No | Unverified | No AI video feature found in TerraMaster material. |
| Surveillance AI | Limited | Unverified | TOS 7 surveillance app supports up to 32 ONVIF IP cameras. Requires TOS 7; no source confirms TOS 7 for this model. |
Calculated from the U24-722-2224'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 | Q4 | Yes | 60GB | Not measured | Estimated |
| Llama 4 Scout 17Bx16ELlama, 109B | None fits | No | 89GB at Q4 | Not measured | Estimated |
| Qwen2.5 72BQwen, 72.7B | Q3 | Yes | 54GB | Not measured | Estimated |
| Mistral Large 2 123BMistral, 123B | Q2 | Yes | 62GB | Not measured | Estimated |
One verdict per workload, from the U24-722-2224's use-case scores and capability rows. Unknown means we cannot confirm it, not that it probably works.
Local LLM use on the U24-722-2224 is OK. Support is limited and Unverified: it means CPU-only inference on the 4-core Xeon via Docker, with no TerraMaster source for this model, although Llama 3.3 70B fits at Q4 (Estimated).
Private RAG is OK, with caveats. RAG itself is limited and Unverified, possible through Open WebUI containers, while Ollama and Open WebUI are vendor confirmed as limited support on x86 TOS NAS without naming this model.
AI search is poor on this system. No local AI or semantic file search is documented for this model, that capability is marked not supported and Unverified, and the Xeon has no NPU to help.
We cannot confirm AI photos on this system. TOS 7 Photos offers face recognition and scene categories, but it requires TOS 7 and no source confirms TOS 7 for this model, so the feature is unconfirmed.
Surveillance support is unknown. The TOS 7 app handles up to 32 ONVIF IP cameras, but it needs TOS 7, which no source confirms here, so camera support is unconfirmed despite the 24 bays for recordings.
The U24-722-2224 is OK for a home lab. It has 24 bays, a Docker manager in TOS and memory that can be expanded to 64GB, though it offers four 1GbE ports and no M.2 slots.
Business use is OK. It offers 24 bays in a rackmount chassis, a 2 year warranty and a 4-core Xeon, though the memory is non-ECC and networking is 1GbE only, so check that against your uptime needs.
Creative AI is unknown on this system. GPU expansion is limited and Unverified, with PCIe 3.0 slots described for 10GbE and RAID cards and no documented GPU use, and acceleration through the integrated UHD Graphics P630 is unconfirmed.
Plan on upgrading. 8GB is fitted as standard, while the U24-722-2224 accepts 64GB of DDR4-3200 non-ECC UDIMM, which is the difference between small models up to around 7B parameters and 32B class models comfortably, and 70B class models at 4-bit with little headroom (Estimated).
Our estimate for the biggest model that fits in 64GB is about 93B parameters at 4-bit, after setting aside 8GB for the system. Long context windows reduce that figure (Estimated).
A graphics card is not an option on the U24-722-2224, which has neither a PCIe slot nor OCuLink. Inference runs on the CPU, with Intel UHD Graphics P630 handling media tasks, so keep to small models.
As the storage half of a RAG setup it works well; as the model host it is limited. Small embedding models run fine on the U24-722-2224, while the answering model needs more compute than it has.
For network access it has 4 x 1GbE. Enough for light use; large imports will take a while.
| Configuration | RAM | SSD | Notes |
|---|---|---|---|
| 8GB | 8GB | None | Diskless; 1x 8GB DDR4 pre-installed, 4 slots. |
The closest TerraMaster model by bay count and AINAS Score. Only the specifications that differ are listed.
Best fit: home lab owners who want one box for storage, containers, media and a few AI services. That is where the U24-722-2224 scores highest (Home Lab 49/100), followed by Business (47/100).
Less suited to buyers who expect to add a graphics card later and teams moving large files over the network every day.
Other AI Capable systems closest in overall score.
Yes, as an estimate. Llama 3.3 70B needs 43GB at Q4 and fits within the 64GB memory ceiling (Estimated), but only after upgrading from the 8GB included. At Q8 it needs 75GB and does not fit (Estimated). This is a calculation, not a benchmark, and inference would be CPU-only, with local LLM support Unverified for this model.
Yes. The memory sits in slots rather than being soldered. It takes DDR4-3200 non-ECC UDIMM, with one 8GB module included across 4 slots and a ceiling of 64GB. The largest module TerraMaster lists is 16GB. Confirm the 64GB ceiling with TerraMaster before buying modules, as an older page quoted a different figure.
No. The memory is DDR4-3200 non-ECC UDIMM, as TerraMaster lists it. The CPU is a Xeon, but the memory the system takes does not provide error correction. Buyers who need ECC for data integrity in business storage should look for a system that lists ECC memory support.
No, not as far as is confirmed. No GPU expansion is listed. The PCIe 3.0 expansion slots are described for 10GbE and RAID cards, the slot count is not confirmed, and GPU use is not documented. The only graphics hardware is the integrated Intel UHD Graphics P630, and there is no NPU.
24. The U24-722-2224 is a rackmount system with 24 bays and no M.2 slots, so all storage sits in the bays. Four 1GbE ports serve those bays, and PCIe 3.0 expansion slots are described for 10GbE and RAID cards, though their count is not confirmed. The 8GB configuration comes diskless, with one 8GB DDR4 module across 4 slots and room to grow to 64GB.
Yes. The system runs TOS, which has a Docker manager on x86 TerraMaster NAS, though TerraMaster does not name this model. Ollama and Open WebUI are vendor confirmed as limited support in containers on x86 TOS NAS generally. Virtual machine support is not confirmed.
Partly. Ollama support is documented by the manufacturer. See the capability table above for the source.
Partly. Open WebUI support is documented by the manufacturer. 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 not yet verified. See the capability table above for the source.
It has 4 x 1GbE built in, and the PCIe 3.0 expansion slots; count not confirmed can take a faster network card. The fastest port is 1GbE.
Every field records its source. Specifications were last checked on . Spot an error? Send a correction with a manufacturer link.