Can it run a 70B-parameter model locally?

Yes. The 128GB of unified memory is sized for large models, and Acer states two linked units can handle models up to 405B parameters. A single unit comfortably covers models in the tens of billions of parameters at reduced precision.

How does it compare with a desktop PC plus a consumer GPU?

A consumer graphics card tops out at 24GB or so of VRAM and needs the rest of the system to feed it over PCIe. The GN100 gives the GPU and CPU one coherent 128GB pool, which is what lets it hold much larger models. It is not a gaming machine.

Is the memory or storage upgradeable?

The 128GB LPDDR5X is soldered unified memory and cannot be expanded. The 2TB NVMe SSD is an M.2 drive, so storage is the part with room to change.

What software comes preinstalled?

It ships with NVIDIA DGX OS and the NVIDIA AI software stack, including PyTorch, Jupyter and Ollama. That means a Linux environment, not Windows.

The workstation itself with its power supply and the preinstalled DGX OS. Check the listing for any included cables or documentation.

Developers, researchers and small teams who need to train or run large models on their own premises and keep data local. Buyers who mainly want gaming or general office work will get more from a standard desktop.

Where is the Acer Veriton GN100 AI Mini Workstation cheapest?

At AAAWave, for $5,999.99, as of our latest check.

What do buyers say about the Acer Veriton GN100 AI Mini Workstation?

Rated 5 out of 5 from 2 ratings.

Acer Veriton GN100 AI Mini Workstation

Best price now

$5,999.99at AAAWaveChecked Nov 11

Prices are checked against each store and labelled with when.

Visit AAAWave

Acer discount code →

Acer Veriton GN100 AI Mini Workstation

Best price now

$5,999.99at AAAWaveChecked Nov 11

Prices are checked against each store and labelled with when.

Visit AAAWave

Acer discount code →

Ask our agent about this product

It checks today's price at every store, compares the alternatives and answers in about a minute.

Price facts

Last seen at $5,999.99 at AAAWave on Nov 11.

Our take on the Acer Veriton GN100 AI Mini Workstation

The Veriton GN100 is a desktop-sized AI workstation built on NVIDIA's GB10 Grace Blackwell Superchip, rated at 1 petaFLOP of FP4 performance. It pairs a 20-core Arm CPU with a Blackwell GPU and 128GB of unified LPDDR5X memory, so large models can run locally without splitting work across separate CPU and GPU memory pools. A 2TB self-encrypting NVMe SSD and a ConnectX-7 SmartNIC support sensitive datasets and linking two units for models up to 405B parameters. It suits developers, researchers and small labs that want local AI compute in a small, quiet box rather than a rack server or a cloud subscription.

Best forAI developers running large models locallyResearch labs and university teamsEdge and office AI deploymentsBuyers who want data kept on-device

Pros

1 petaFLOP FP4 from the GB10 Grace Blackwell Superchip
128GB unified memory shared coherently by CPU and GPU
2TB self-encrypting NVMe storage for sensitive data
ConnectX-7 networking allows two-unit scaling to 405B models
Ships with DGX OS and a ready AI stack (PyTorch, Jupyter, Ollama)
Very small footprint at 150×150mm and under 1.5kg

Cons

Arm-based platform, so some x86 software needs porting or emulation
Unified LPDDR5X memory is soldered and cannot be upgraded
No discrete GPU slot or user-expandable memory
NVIDIA DGX OS is Linux-only, with a learning curve for Windows users
ConnectX-7 and dual-unit scaling assume a wired enterprise network

Acer Veriton GN100 AI Mini Workstation prices at every store

AAAWaveNew
In stockReturn policy
$5,999.99Checked Nov 11
Visit AAAWave

About the Acer Veriton GN100 AI Mini Workstation

The Acer Veriton GN100 AI Mini Workstation (VGN100-UD13) is a compact desktop AI computer built around the NVIDIA GB10 Grace Blackwell Superchip. It delivers up to 1 petaFLOP of FP4 AI performance in a 150×150mm chassis that weighs under 1.5 kg. The chip pairs a 20-core ARM CPU with a Blackwell GPU that has 5th-generation Tensor Cores and 4th-generation RT cores.

It has 128GB of LPDDR5X unified memory on a 256-bit bus, shared between the CPU and GPU, for running large language models, fine-tuning and multi-modal AI workloads locally. A 2TB self-encrypting NVMe SSD stores datasets and model checkpoints.

The NVIDIA ConnectX-7 SmartNIC lets two units be linked for multi-node scaling, supporting models up to 405 billion parameters. NVIDIA DGX OS comes pre-configured with PyTorch, Jupyter, Ollama and NVIDIA developer tools. Wi-Fi 7, Bluetooth 5.4 and a Kensington lock slot are included.

Model: VGN100-UD13
Part number: DT.R6LAA.003
SoC: NVIDIA GB10 Grace Blackwell Superchip
CPU: 20-core ARM (10× Cortex-X925 + 10× Cortex-A725)
GPU: NVIDIA Blackwell with 5th-gen Tensor Cores and 4th-gen RT Cores
AI performance: Up to 1 petaFLOP (FP4)
Memory: 128GB LPDDR5X, 256-bit bus (8×16GB)
Storage: 2TB NVMe M.2 SSD, self-encrypting
Networking: NVIDIA ConnectX-7 SmartNIC, Gigabit Ethernet
Wireless: Wi-Fi 7 (802.11be), Bluetooth 5.4
Ports: 4 x USB, HDMI
Power supply: 170W
Operating system: NVIDIA DGX OS
Software: PyTorch, Jupyter, Ollama, NVIDIA AI tools
Dimensions: 5.9" × 5.9" × 2.0" (W×D×H); 150×150mm footprint
Weight: Under 1.5 kg
Color: Black
Security: Kensington lock slot, self-encrypting SSD
Multi-unit scaling: Two units linked via ConnectX-7, up to 405B parameter models
Pack options: Pack of 1; Pack of 2 with cable (DT.R6LAA.003 X2+CABLE). The included cable is a compatible replacement, not an original manufacturer cable.

Questions and answers

Can it run a 70B-parameter model locally?: Yes. The 128GB of unified memory is sized for large models, and Acer states two linked units can handle models up to 405B parameters. A single unit comfortably covers models in the tens of billions of parameters at reduced precision.
How does it compare with a desktop PC plus a consumer GPU?: A consumer graphics card tops out at 24GB or so of VRAM and needs the rest of the system to feed it over PCIe. The GN100 gives the GPU and CPU one coherent 128GB pool, which is what lets it hold much larger models. It is not a gaming machine.
Is the memory or storage upgradeable?: The 128GB LPDDR5X is soldered unified memory and cannot be expanded. The 2TB NVMe SSD is an M.2 drive, so storage is the part with room to change.
What software comes preinstalled?: It ships with NVIDIA DGX OS and the NVIDIA AI software stack, including PyTorch, Jupyter and Ollama. That means a Linux environment, not Windows.
What is in the box?: The workstation itself with its power supply and the preinstalled DGX OS. Check the listing for any included cables or documentation.
Who is it for?: Developers, researchers and small teams who need to train or run large models on their own premises and keep data local. Buyers who mainly want gaming or general office work will get more from a standard desktop.
Where is the Acer Veriton GN100 AI Mini Workstation cheapest?: At AAAWave, for $5,999.99, as of our latest check.
What do buyers say about the Acer Veriton GN100 AI Mini Workstation?: Rated 5 out of 5 from 2 ratings.