⚡ggml/llama.cpp has been ported to run on a jailbroken PlayStation 5, enabling native execution of GGUF models on the console.

"Llama.cpp can now run on a jailbroken PS5" — Hacker News (best) · AI & Technology

17 Oct 2026

What it means

This is a project called PS5LM that ports — the C++ inference engine for LLaMA-style models — to a PlayStation 5. It lets the console run -format AI models natively using the PS5's GPU and additional RAM, starting with models like 3.8. Reports indicate it can handle models up to 27 billion parameters at speeds over 100 tokens per second, effectively turning a gaming console into a local AI inference server.

Accuracy & sourcesGrounded

The claim is solid: the primary source is the open-source GitHub repository cobanov/PS5LM, which contains the code and documentation. Tech outlets like Tom's Hardware and Wccftech have verified and reported on the project's capabilities, and the repository itself demonstrates functional inference.

⚡ggml/llama.cpp has been ported to run on a jailbroken PlayStation 5, enabling native execution of GGUF models on the console.

What it means

Accuracy & sourcesGrounded

How new