A new GitHub project called PS5LM lets a jailbroken PlayStation 5 run llama.cpp natively, opening the door to local large language models on Sony's console. The repository, published by developer cobanov, states that any GGUF model can be loaded on the system, with Qwen 3.8 listed as the first supported model.
The project does not use the PS5's game operating system or streaming from a PC. According to the repository description, it runs llama.cpp directly on the console hardware, which is why the story is spreading quickly among gamers and AI enthusiasts on social platforms.
PS5LM and GGUF support
The repository is organized into folders for docs, patches, probes and a dedicated ps5 directory, suggesting a structured port rather than a one-off script. The README explicitly mentions Qwen 3.8 as the starting point, which points to a focus on the Qwen family of open-weight models.
GGUF is the file format used by llama.cpp to package model weights and metadata. Because the project claims support for any GGUF model, users could in theory load different sizes and quantizations, although the PS5's memory and the jailbreak environment will set the real limits.
- Project name: PS5LM, hosted on GitHub by cobanov
- Runtime: llama.cpp running natively on a jailbroken PS5
- Model format: GGUF
- First model mentioned: Qwen 3.8
- Repository folders: docs, patches, probes, ps5
