Lenovo has officially introduced the ThinkStation PGX, a compact yet powerful AI developer workstation that delivers data center-level performance in a desktop-friendly form factor. Designed for AI researchers, developers, data scientists, and students, the PGX is powered by the NVIDIA GB10 Grace Blackwell Superchip, offering up to 1 PetaFLOP (1,000 TOPS) of AI performance. This remarkable processing capability allows users to work with massive generative AI models containing up to 200 billion parameters.
The workstation is equipped with 128 GB of unified system memory, ensuring efficient handling of large-scale AI workloads, model experimentation, fine-tuning, and inference. For more demanding projects, two PGX systems can be connected to support models with up to 405 billion parameters, providing significant scalability.
Out of the box, the ThinkStation PGX comes preloaded with NVIDIA’s DGX OS and AI software stack, including popular tools and frameworks such as PyTorch and Jupyter, enabling seamless development and testing directly from the desktop. Thanks to the consistent software architecture of NVIDIA’s AI platform, developers can easily deploy their models from the PGX to larger data centers or cloud environments with minimal code modifications. Set to launch in Q3 2025, the ThinkStation PGX positions itself as a premium solution for professionals and organizations seeking high-performance, localized AI development capabilities.
Lenovo has confirmed that the PGX will be released in the third quarter. While the company has not yet announced official pricing, it is expected to fall within the $3,000 to $4,000 range — similar to the ASUS and NVIDIA GB10 systems.
Source : Lenovo