PSllama is an open‑source project that runs LLMs, vision, voice and image‑generation models entirely on the PlayStation 5’s GPU. It bypasses the CPU by hooking into the console’s RDNA2 architecture through the Mesa RADV Vulkan driver from the PS5_Vulkan effort. The implementation combines llama.cpp, stable‑diffusion.cpp and whisper.cpp into a unified framework, with source code and interactive demos available on GitHub.