AI WorkStation - Lenovo ThinkStation PGX

1.13-litre AI workstation on the NVIDIA GB10 Grace Blackwell Superchip: 20-core Grace Arm CPU, Blackwell GPU, 1 PFLOP FP4, 128GB unified memory, 4TB NVMe.

€6,000
In stock Order this system

1-year limited warranty and technical support. Delivery 14-21 business days. Custom upgrades available on request.

Please note

Laptop, monitor and wireless keyboard shown in the gallery are optional and sold separately.

Did you know?

ThinkStation PGX is for AI developers, data scientists, students, researchers, and AI enthusiasts: anyone seeking scalable, local inferencing without cloud complexity and at a fraction of the cost of cloud resources.

Lenovo Hybrid AI Advantage with NVIDIA

Lenovo is at the forefront of delivering AI-ready performance across its workstation portfolio, from powerful ThinkStation desktops to mobile ThinkPad P Series workstations. Through Lenovo's Hybrid AI Advantage with NVIDIA, Lenovo offers full-stack, validated AI solutions that accelerate the development and deployment of agentic AI: systems capable of reasoning, planning, and acting to achieve goals efficiently. These scalable solutions empower enterprises to harness AI across diverse environments, from edge computing to cloud infrastructures, enhancing productivity and driving innovation in the AI era.

Lenovo ThinkStation PGX is the perfect companion for Lenovo's Hybrid AI platforms:

  • Hybrid AI 285 platform - based on the NVIDIA 2-8-5 PCIe-optimised configuration (2x CPUs, 8x GPUs, and 5x network adapters), ideally suited for medium (per GPU) to large (per node) inference use cases.
  • Hybrid AI 289 platform - based on the NVIDIA 2-8-9 configuration (2x CPUs, 8x GPUs, and 9x network adapters), ideally suited for model training and large-scale inference use cases.
  • ThinkStation PGX provides developers with a powerful, economical platform for prototyping models and AI applications, freeing up valuable computing resources in on-prem cluster and/or cloud compute environments that are better suited to training and/or deploying production AI models.
  • Leveraging the NVIDIA AI platform software architecture makes it possible for ThinkStation PGX users to seamlessly move their models from their desktop to any accelerated cloud or data centre infrastructure with virtually no code changes, making it easier than ever to prototype, fine-tune, and iterate.

Scalable AI development - AI power, perfectly packaged

The Lenovo ThinkStation PGX small form factor, powered by the NVIDIA GB10 Grace Blackwell Superchip, enables simplified AI workflows right at your desk. It comes preloaded with the NVIDIA AI software stack and seamlessly integrates with your workstation for on-device AI development, eliminating the complexity and cost of cloud-based solutions.

Next-gen Blackwell GPU - Seamless integration into AI pipelines

An advanced NVIDIA Blackwell GPU supports complex AI workflows, including fine-tuning, inference, and data processing, with features designed to streamline deep learning, computer vision, and generative AI development.

AI excellence - Big models, small chip, limitless potential. The NVIDIA GB10 Grace Blackwell Superchip combines cutting-edge integrated graphics with a powerful multicore Arm CPU to deliver an energy-efficient system-on-a-chip (SoC) solution. It's built to handle complex AI tasks and massive datasets with incredible speed and efficiency, making it perfect for tackling demanding workloads locally.

Full-stack NVIDIA AI platform - Ready-to-go AI development environment. A compact entry point to advanced AI, the ThinkStation PGX is preloaded with the NVIDIA DGX™ OS and NVIDIA AI software stack, providing an optimised space for AI model development. With tools like PyTorch and Jupyter® Notebooks, engineers can easily prototype, fine-tune, and inference models for smooth deployment to data centre or cloud platforms.

Sandboxed development environment - Your AI playground: experiment, build, succeed. The ThinkStation PGX workstation offers developers a secure sandbox for prototyping AI models, freeing up cluster resources for production workloads. This streamlines workflows, keeps IP on-premises, and reduces security risks: ideal for experimentation without compromising enterprise standards.

  • Tools and frameworks: The ThinkStation PGX comes preconfigured with the NVIDIA DGX OS and the NVIDIA AI software stack, along with familiar tools and frameworks like PyTorch and Jupyter. Developers can instantly prototype, fine-tune, and inference large AI models from the desktop and seamlessly deploy to the data centre or cloud.
  • Built on the NVIDIA GB10 Grace Blackwell Superchip: providing up to 1 PetaFLOP (1000 TOPS) of AI performance, the ThinkStation PGX can tackle large generative AI models of up to 200 billion parameters. With 128GB of coherent unified system memory, developers can experiment, fine-tune, or inference the latest generation of reasoning AI models. To double down on computing power, developers can connect two ThinkStation PGX systems together to work with even larger AI models up to 405 billion parameters.
  • Powerful, economical platform for prototyping models and AI applications: ThinkStation PGX provides developers with a powerful, economical platform for prototyping models and AI applications, freeing up valuable computing resources in on-prem cluster and/or cloud compute environments that are better suited to training and/or deploying production AI models. Leveraging the NVIDIA AI platform software architecture makes it possible for ThinkStation PGX users to seamlessly move their models from their desktop to any accelerated cloud or data centre infrastructure with virtually no code changes, making it easier than ever to prototype, fine-tune, and iterate.
  • Powered by the NVIDIA Grace Blackwell GB10 Superchip: The Lenovo AI Workstation ThinkStation PGX is powered by the NVIDIA Grace Blackwell GB10 Superchip and features 128GB of unified system memory, a ConnectX-7 SuperNIC and more. It is capable of handling up to 200B parameter LLMs (405B if you connect two together with the optional cable kit). Lenovo designed it to be YOUR AI appliance, YOUR personal AI cloud, a device that will enhance YOUR efficiency and YOUR own AI development needs. No more waiting in line for server resources, no more cloud credits, no more wasted cycles trying to locate suitable GPU instances to support your needs... just AI productivity and agility at YOUR fingertips. Use standalone, as a pair for enhanced capabilities, or operate remotely from your own laptop. Your perfect companion for 2026+.

System specifications

Architecture
NVIDIA Grace Blackwell
Processor
NVIDIA Grace 20-core Arm® CPU (10 Cortex-X925 cores, 10 Cortex-A725 cores)
Graphics
NVIDIA Blackwell architecture: CUDA® cores, 5th gen Blackwell Tensor Cores, 4th gen RT Cores, dedicated NVENC video encoder
AI performance
1000 TOPS, 1 PFLOP (FP4, sparsity)
Unified system memory
128GB LPDDR5x, 256-bit bus, 273 GB/s bandwidth
Total storage
4TB NVMe M.2 with self-encryption
Scalability
Two ThinkStation PGX nodes can be linked via integrated NVIDIA ConnectX®-7 to handle models up to 405B parameters
Power consumption
240W
Installed software
NVIDIA AI software stack, CUDA 13, GB10 Dashboard, AI Workbench
Rear ports
1 x USB-C for power (PD 3.1); 3 x USB-C (USB4 20Gb/s) with DisplayPort 2.1 support; 1 x HDMI 2.1a (multichannel audio); 1 x RJ-45 LAN 10GbE; 2 x QSFP ports for NVIDIA ConnectX-7 (required to connect to other ThinkStation PGX units; cables sold separately)
Wi-Fi
Wi-Fi 7 and Bluetooth® 5.3 LE
Security
Self-encrypting NVMe (AES SED), TPM 2.0, NVLink-C2C enclave, NVIDIA FW Recovery, AMI setup password, UEFI Secure Boot
Design
1.13 litre (small form factor)
Colour
Dark grey
Dimensions
150 x 150 x 50.50 mm (5.91 x 5.91 x 1.99 inches) (W x D x H), 1.2 kg (2.65 lbs)
Usage
AI development and programming: AI developers, data scientists, students, researchers, and AI enthusiasts
Warranty
1 year warranty

Order this system

Send us your details and we will confirm availability, final price and delivery before anything is charged. Payment by card is available through the secure Stripe checkout button above once enabled; bank transfer details are shown alongside.