





Memory decides whether a model fits on the machine at all; bandwidth decides how fast it answers. The selected configuration has 96GB of memory and 1.2TB/s of bandwidth.
| Memory | Available to the model | Max model size | What it can do |
|---|---|---|---|
| 36GB | ~27GB | ~43B parameters | Q&A over internal documents, summarizing, drafting — small teams |
| 48GB | ~36GB | ~57B parameters | Q&A over internal documents, summarizing, drafting — small teams |
| 64GB | ~48GB | ~76B parameters | Larger models, longer context, serving many concurrent users |
| 96GB | ~72GB | ~115B parameters | Larger models, longer context, serving many concurrent users |
| 128GB | ~96GB | ~153B parameters | Very large models, or several models side by side on one machine |
| 256GB | ~192GB | ~307B parameters | Very large models, or several models side by side on one machine |
Estimates based on: a 4-bit quantized model takes ~0.5 bytes per parameter; macOS lets the GPU use ~75% of memory; 20% is reserved for context. These are not benchmark results — we test on real hardware for you before you commit.
Up to 48GB · your 96GB configuration
Q&A over your internal documents, summarizing and drafting for a small group of users. A good fit if your business is new to AI and wants its data to stay in the office.
64–96GB · your 96GB configuration
Larger models and longer context (multi-page files and contracts), serving many people asking questions on the same internal network.
128GB and up · your 96GB configuration
Run very large models, or several models side by side on one machine (for example, one model for Q&A and one dedicated to document processing).
There is no fixed user count per tier: it depends on the model, prompt length and how many people ask at once. Namtech measures against your real workload before you commit to a configuration.
| Chip | Apple M5 Ultra |
|---|---|
| CPU | 36 cores |
| GPU | 80 cores |
| Neural Engine | 32 cores |
| Unified memory | 96GB |
| Memory bandwidth | 1.2TB/s |
| Storage | 1TB |
| Ports | 4× Thunderbolt 5 (USB-C, up to 120Gb/s, DisplayPort 2.1) · 2× USB-A (up to 5Gb/s) · 1× HDMI 2.1 · 1× 10Gb Ethernet (Nbase-T, RJ-45) · 3.5 mm headphone jack · Front: SDXC card slot (UHS-II) + 2× USB-C 10Gb/s (M5 Max) or 2× Thunderbolt 5 (M5 Ultra) · Wi-Fi 7 (802.11be), Bluetooth 6, Thread — Apple N1 chip |
| Display support | M5 Max: up to 5 external displays · M5 Ultra: up to 8 external displays (6K@60Hz) |
| Dimensions | Height 9.5 cm · Width 19.7 cm · Depth 19.7 cm |
| Weight | 3.6 kg |
| Power | 100–240V AC, maximum continuous power 480W · operating temperature 10–35°C |
| In the box | Mac Studio and power cord |
Specifications from apple.com/vn, updated September 24, 2026. Prices on this page include VAT.
Genuine Apple keyboards, mice, trackpads, cables and Studio Display monitors compatible with Mac Studio and Mac mini, selected using the compatibility filter on apple.com/vn. You can order them together with your Mac.
























About 115B parameters at 4-bit quantization. With 96GB of unified memory, about 72GB is available for the model after leaving room for the operating system and conversation context. This is an estimate to help you choose a machine; the real figure depends on the specific model, context length and number of concurrent users.
Yes. The 96GB configuration holds models up to about 115B parameters at 4-bit quantization, so a 70B model fits entirely on the machine. If you want to run 70B at lighter quantization for higher-quality answers, or with many people asking at once, consider a version with more RAM.
Many users, larger models. Larger models and longer context (multi-page files and contracts), serving many people asking questions on the same internal network. The actual number of concurrent users depends on the model and prompt length, so Namtech measures on real hardware against your workload before you commit to a configuration.
234.900.000₫, VAT included, with the default 1TB SSD. Namtech issues invoices to businesses and advises on the configuration for the models you need to run. Storage can be upgraded from 1TB to 16TB, with the price updating instantly in the configurator at the top of the page.
No. Mac Studio memory is built into the chip, so you must choose the right amount when you order. That is why you should work out the size of the models you plan to run before you decide.
Choose 96GB if the model you need to run is under about 115B parameters. The Mac Studio M5 Ultra 36CPU 80GPU 256GB 1TB runs up to about 307B, priced at 366.900.000₫ (132.000.000₫ more), VAT included. Since RAM cannot be upgraded after purchase, choose the next tier up if you are unsure.
Tell us what you plan to use AI for and how many people will use it — Namtech will recommend a configuration that fits your needs, with no overselling.