Developer Turned PlayStation 5 Into AI Server

The new PS5LM project enables jailbroken consoles to run large language models using their built-in GPU hardware.

Updated on Oct. 14, 2026 in Artificial Intelligence

Isometric editorial illustration of a gaming console mainboard and silver heat-sink assembly floating in a clean industrial void.
The new PS5LM project enables jailbroken PlayStation 5 consoles to function as local AI servers by leveraging their high-bandwidth GPU memory. AI Illustration. Upload story photo >

Live Poll

Would you modify your home gaming console to run advanced AI models?

A developer named cobanov has released the PS5LM project, which allows jailbroken PlayStation 5 consoles to function as local AI servers. The tool utilizes the console's AMD GPU to run advanced language models efficiently.

Why it matters

The project exploits the high memory bandwidth of the PlayStation 5, demonstrating that consumer gaming hardware can serve as a viable platform for running resource-intensive large language models.

The PS5 hardware features an AMD GPU with 16GB of GDDR6 memory and 448 GB/s of memory bandwidth. It supports models like Qwen3.8-27B at over 20 tokens per second via a ported llama.cpp engine and a RADV-based Vulkan driver.

The players

cobanov

This individual is the developer who created and released the PS5LM project.

Alibaba

This is the multinational technology company that originally developed the Qwen series of language models.

The details

By leveraging the relapse browser exploit on firmware versions 7.00 through 13.60, the PS5LM project allows users to bypass standard restrictions to run AI inference. The software routes calculations through the GPU, which allocates 11.44 GiB of device heap for model operations.

Timeline

  1. August 14: Alibaba released the Qwen3.8-27B model.

  2. October 14, 2026: The PS5LM project was publicly announced.

The Tech Race

This project extends the utility of the llama.cpp inference engine by successfully porting it to the PlayStation 5 gaming console. It highlights a growing trend of repurposing consumer-grade gaming hardware to bypass the high costs of specialized enterprise AI servers.

Users with jailbroken consoles can now run private, local AI models without relying on cloud services or expensive GPUs. However, adopting this project requires specific firmware versions and carries the inherent risk of voiding hardware warranties.

The takeaway

This development proves that powerful AI performance is increasingly accessible through common home hardware. Enthusiasts can leverage this project to gain local control over data processing while avoiding subscription-based cloud AI tools.

Further reading

Learn more about the latest developments in Artificial Intelligence.

Live Poll

Would you modify your home gaming console to run advanced AI models?