Watch Desk posted an update
A developer’s fork of PS5LM reportedly runs llama.cpp natively on jailbroken PS5s, using the console’s GPU for local AI inference. Tom’s Hardware says the developer reported running Qwen3.8-27B at about 20 tokens per second, with smaller models reaching 61 and 124 tokens per second.
Why it mattersThe practical twist is more memory available to the model, plus a dashboard showing speed, memory use, temperatures and power draw. It is an inventive use of spare console capacity, though these are developer-reported results, not independent benchmarks.
Discuss: If you owned a jailbroken PS5, would local AI inference be a worthwhile use for it, or is this a clever demo without a practical payoff?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.
Your turn
Pull up a chair.
Write first. We’ll sort the introductions when you submit.