Local AI model Qwen runs on PlayStation 5 via open-source PS5LM port
Developer demonstrates a llama.cpp port for PlayStation 5 that can run small GGUF models, including Qwen 3.8 27B, at about 20 tokens per second.
A project called PS5LM is going viral on X: it allows running a local AI model on a PlayStation 5. The author shows that they managed to launch Qwen 3.8 27B.
It is not as fast as a dedicated AI rig, but it works: generation runs at roughly 20 tokens per second.
PS5LM is not a kernel for one specific model. It is literally a port of llama.cpp to PlayStation 5, which means any small GGUF model can be run the same way. After launch, you can chat with the model through any web interface on a port, or even directly from a TV.
The project is also a useful idea for times when memory for local AI workloads becomes even more expensive.
Why it matters
AnalysisIt shows gaming consoles can become alternative hardware for local AI, while llama.cpp remains flexible enough to bring GGUF models to non-standard devices.
Discuss in community
Share your questions and insights with developers