Jailbroken PS5s can now run local AI models at over 100+ tokens per second

PS5LM gets a fork, allowing it to run natively and use the GPU and more RAM to run larger LLMs, even running Qwen3.8-27B at 20 tokens per second.