PS5LM enables local GGUF model serving via jailbroken PS5 GPU
October 16, 2026
PS5LM uses llama.cpp with a Vulkan backend to serve GGUF models from a jailbroken PS5's RDNA 2 GPU. It provides 11.4 GB of shared memory for context and supports OpenAI and Anthropic compatible APIs for local network use.
HOW THIS AFFECTS YOU
●
builderYou can use consumer hardware for private, local model inference serving.