You can run a Qwen 3.8-27B on a PS5.
Qwen 3.8-27B runs on a jailbroken PS5 at 21 tokens per second
AISummary
A project called PS5LM runs Qwen 3.8-27B on a jailbroken PS5 using the console's GPU, with heavy compression to fit the 16GB of memory. The model generates about 21 tokens per second, while smaller models reach 61 tokens per second with Gemma 4 E4B and 124 with Llama 3.2 3B.
Post on XView on X
Andrew CurranVerified on X
@AndrewCurran_
A jailbroken PS5 is now being used to run a powerful AI model directly on the console. A project called PS5LM has managed to run Qwen 3.8 27B, an AI model with around 27 billion parameters, using the PS5’s GPU. The PS5 has only 16GB of memory, so the AI model needs heavy compression to fit, even with that limitation, it can generate around 21 tokens per second. Smaller AI models are much faster. Tests reached around 61 tokens per second with Gemma 4 E4B and up to 124 tokens per second with Llama 3.2 3B. During the test, the PS5’s GPU was working at 100%, while the CPU was barely being used. The console used around 86W of power and reached about 54°C.
Source: Andrew Curran · x.comPublished · added here