Garlic LLM runs Qwen3.5-9B AI model on jailbroken PS4 hardware
Developer Leonardo Alegría released Garlic LLM, an open-source app that lets a jailbroken PS4 run Alibaba's 9-billion-parameter Qwen3.5-9B model entirely offline, using the console's 2013-era GPU. The model is compressed to roughly 4.5GiB and runs at 31-39 tokens per second, with voice chat and dictation features powered by additional speech models. It requires a PS4 running the GoldHEN homebrew enabler and is distributed as a free 5.66GB download via Hugging Face.
GoKawiil's interpretation of the reporting above, not reported fact.
The project demonstrates that modest, decade-old console GPUs can run modern mid-sized language models without relying on existing graphics driver layers like Vulkan, which could inform future homebrew or low-power AI deployment efforts. It also extends a growing trend of repurposing gaming consoles as local AI inference devices, following similar work on PS5 hardware, suggesting enthusiast developers see untapped potential in fixed-function console silicon for offline AI tasks.
- Garlic LLM runs the 9B-parameter Qwen3.5-9B model on a jailbroken PS4 at 31-39 tokens per second.
- The app requires GoldHEN homebrew and works without llama.cpp or a Vulkan driver port, unlike the PS5LM project.
- It supports voice conversation in seven languages and dictation in 25 European languages via added speech models.
Sony PlayStation 4 Console — If you're curious about running homebrew AI projects like Garlic LLM, you'll need a PS4 to experiment with, and the base/Slim model is exactly what the developer used for testing. It's a great way to repurpose older hardware for fun offline AI experiments while still having a capable gaming console.
See Sony PlayStation 4 Console on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: tomshardware.com — Shane Downing, 2026-11-16
Published there as: “Jailbroken PS4 runs 9-billion-parameter AI model at up to 43 tokens per second”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.