Lenovo has introduced the ThinkStation PGX, a new compact desktop workstation purpose-built for artificial intelligence (AI) development and research. This small-form-factor device is designed to meet the needs of AI researchers, developers, data scientists, practitioners, students, and application engineers seeking a powerful yet space-efficient solution that is ready to use out of the box. At the heart of the ThinkStation PGX is the NVIDIA GB10 Grace Blackwell Superchip, which delivers up to 1 PetaFlop (1000 TOPS) of AI performance, enabling the system to efficiently handle large generative AI models containing up to 200 billion parameters.
Developers benefit from 128 GB of unified, coherent system memory, allowing for robust experimentation, fine-tuning, and inference with advanced reasoning AI models. The workstation also offers scalable performance: by linking two ThinkStation PGX units, users can tackle even larger AI models with up to 405 billion parameters. This modular approach helps ensure that the platform can grow alongside evolving artificial intelligence workloads, supporting cutting-edge research and application development.
The ThinkStation PGX ships preconfigured with NVIDIA DGX OS and the NVIDIA artificial intelligence software stack, providing seamless access to popular development tools and frameworks like PyTorch and Jupyter. This enables developers to rapidly prototype, refine, and infer large AI models directly from the desktop, while offering streamlined deployment to data centers or cloud environments as workflows mature. According to Rob Herman, Vice President, Worldwide Workstation and Client AI Business at Lenovo, the collaboration with NVIDIA is aimed at empowering the next generation of AI practitioners by making high-end, compact solutions more accessible to individual developers, researchers, and students alike.
