Dell unveils Pro Max with GB10 for on-site large-scale AI work
Dell has announced the release of the Dell Pro Max with GB10, a desktop system tailored for local AI development and designed to handle large-scale models directly at the point of work.
The Dell Pro Max with GB10 is described as the first of its kind in the Dell Pro Max portfolio, specifically targeting organisations and individuals developing and implementing artificial intelligence models that exceed the capacity of traditional high-end workstations.
Hardware and performance
Purpose-built for deskside AI prototyping, the Dell Pro Max with GB10 features a compact form factor powered by the NVIDIA Grace Blackwell Superchip. According to Dell, the system supports up to 128GB of LPDDR5x unified memory, facilitating local processing of models with up to 200 billion parameters. Its computing power reaches up to 1000 TFLOPS of FP4 performance.
The system runs NVIDIA DGX OS and the complete NVIDIA AI software stack, ensuring consistency for developers who want to move models between desktop and Dell data centre environments. Out-of-the-box software includes CUDA, JupyterLab, Docker, and AI Workbench. Dell states that these pre-installed tools allow teams to begin prototyping and building as soon as the system is unboxed.
A change in AI development
Charlie Walker, representing Dell's technical communications, noted the context behind the product's development and what the company believes it offers compared to both existing desktop systems and remote cloud services.
Human ingenuity fuels AI progress, yet most teams hit hard limits on computation well before reaching their creative potential. I continue to hear from customers determined to solve tough challenges, only to run into the current threshold of high-end workstations that max out at 45 billion parameters-even our highest-spec Dell Pro Max with 96GB of GPU VRAM. When memory becomes the bottleneck, teams face a choice: compress models and risk sacrificing accuracy or lean on expensive cloud solutions that are rarely the best fit for every need. For regulated industries, moving sensitive data off-premises isn't even on the table.
Walker continued,
That's why the arrival of Dell Pro Max with GB10 marks a notable shift. Powered by NVIDIA Grace Blackwell and running DGX OS, this desktop system delivers true data center-class AI performance at the point of work. It features 128GB unified memory to support models up to 200 billion parameters and can reach 1000 FP4 TOPS-a rare leap in local capability. With pre-installed tools like CUDA, JupyterLab, Docker and AI Workbench, teams can unbox and start building in minutes. For those needing even more power, connecting two Dell Pro Max with GB10 systems turns them into a single node capable of supporting 400 billion-parameter models, a clear demonstration of Dell's approach to scalable AI infrastructure.
The company outlined several user groups that may benefit from the GB10 system, including academic researchers, startups, regulated industries, and individual creators.
Researchers and startups
For academic research, Dell claims that the system allows direct experimentation with large models like Llama 3.3 70B without relying on remote resources, expediting investigations and reducing waiting times for access to high-powered computing. In startup environments, the unified memory architecture is intended to allow for local prototyping, fine-tuning, and scaling of AI models, reducing reliance on cloud infrastructure and its associated costs or complexities.
In Walker's words,
Startups face daily trade-offs between building products and investing in the infrastructure needed to support growth. Dell Pro Max with GB10 addresses this by allowing founders to inference, fine-tune, prototype, validate, and scale AI models locally. Its unified memory architecture eliminates the complexity often seen in distributed solutions, so teams stay focused on solving real problems-not wrestling with setup or unpredictable costs. And a two-person team can now tap into computational resources that were previously within reach only for enterprises able to invest heavily in infrastructure.
Regulated sectors and individual creators
For regulated industries such as banking and healthcare, Dell states that the new hardware allows them to retain data on premises, which could enable the training and deployment of sensitive AI applications including fraud detection and medical imaging while adhering to privacy requirements.
Walker commented,
For banks and healthcare providers, the ability to keep sensitive data on-site is non-negotiable. Dell Pro Max with GB10 enables these organizations to train and deploy advanced AI-such as fraud detection or diagnostic imaging-entirely within their own environment. Medical IT teams have shared that this approach delivers performance on par with leading cloud solutions, all while protecting privacy and reducing risk.
He also said that individual creators now have access to a level of computation that was previously unattainable without large-scale cloud access:
The same power that drives enterprise AI is now available to independent developers, creators, and entrepreneurs. The Dell Pro Max with GB10 makes sophisticated AI development accessible to individual creators who once lacked computational resources to bring them to life. Where hardware or cloud limitations once forced innovation to slow, Dell Pro Max with GB10 puts data center capability on the desktop. I've seen creators fine-tune vision models for client portfolios and game developers train custom AI characters-all from their own workspace, with no need for external infrastructure or added costs.
Deployment and availability
The Dell Pro Max with GB10 is presented as scaleable with the possibility of linking two units for even larger models. Dell summarised the product's purpose as, "an AI accelerator for real-world AI challenges, built for those who won't let limits define what's possible."
