Lenovo ThinkStation PGX AI Workstation · NVIDIA GB10 Grace Blackwell · 128GB Unified Memory · 4TB SSD · NVIDIA DGX OS

The Lenovo ThinkStation PGX (30KL000AGF) is a palm-sized AI workstation built on the NVIDIA GB10 Grace Blackwell Superchip, with 128GB of unified LPDDR5X memory and a 4TB PCIe 5.0 SSD, made for developers, data scientists and researchers who want to prototype, fine-tune and run large AI models locally.

It runs NVIDIA DGX OS with the NVIDIA AI software stack, delivers up to 1 petaFLOP of FP4 AI performance according to Lenovo, and connects over 10GbE, Wi-Fi 7 and two 200Gbps QSFP ports from its ConnectX-7 network chip. Lenovo states that one unit handles models of up to about 200 billion parameters, and two linked units up to about 405 billion.

• ThinkStation PGX unit
• 240W USB-C power adapter
• No keyboard, mouse or monitor are listed for this model
• The QSFP cable for linking two units is sold separately (Lenovo PN 4X91U42988)

• 128GB of unified memory is shared by the 20-core Arm CPU and the Blackwell GPU, so large models that would not fit in a typical graphics card's memory can be loaded and tested on the desk.
• The 4TB self-encrypting (Opal) PCIe 5.0 SSD gives room for datasets, model checkpoints and containers.
• Two ThinkStation PGX units can be linked through their 200Gbps QSFP ports to work on larger models; Lenovo notes that more units need a QSFP switch.
• At 150 x 150 x 50.5 mm and 1.2 kg it sits next to a laptop or monitor and can be reached remotely over the 10GbE network.

• Lenovo does not offer Windows for this system: it ships with NVIDIA DGX OS on an Arm CPU, so x86-only Windows applications are not an option.
• There are no PCIe slots and memory is soldered, so RAM and GPU cannot be upgraded.
• No front ports — all connections are on the rear panel.
• Not intended for gaming or as a Windows office PC.