TL;DR
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
PS5LM is a homebrew app that runs llama.cpp on a jailbroken PlayStation 5 and serves local GGUF models to other devices on the home network. Its developer reports model-speed and decision-model tests on a PS5 Slim, but results may vary; the app requires a jailbroken console running firmware 7.00 to 13.60.
PS5LM, a homebrew project for jailbroken PlayStation 5 consoles, runs llama.cpp on the console’s GPU and serves local GGUF language models to devices on the same network. The project says the setup can respond to requests from laptops, phones and coding tools without a separate PC or cloud service, but it requires a compatible jailbroken console.
The project says it uses llama.cpp’s Vulkan backend with the PS5’s RDNA 2 GPU, assigning up to 11.4 GB of shared memory to a model and its context. It provides a server at a local network address and supports OpenAI-compatible chat completions and Anthropic-compatible messages, as well as a browser-based chat page. Those interfaces let existing clients direct requests to the console rather than a hosted service.
PS5LM’s release page reports tests on a PS5 Slim: Qwen 3.5 0.8B at 4-bit quantization produced 207 tokens per second; Qwen 3.5 9B at 4-bit produced 55 tokens per second; and Qwen 3.8 27B at a listed UD-Q2_K_XL quantization produced about 20 tokens per second with a 64,000-token context. These are figures reported by the project, not independently verified benchmarks, and the source does not provide a standardized comparison method.
The app also supports decision models with llama.cpp decision metadata. PS5LM says these models can return structured choices, scores or yes-or-no answers without generating a conventional text response. Its example, using the d1 Omni 600M model, reports three questions answered in 0.3 seconds and a test involving a 720p image and seven questions completed in 2.9 seconds. The project also describes an on-TV interface for loading models, viewing requests, and monitoring console temperature and hardware use.
A Console Becomes a Local Model Server
The project repurposes a game console as a local inference server, allowing multiple devices in a household to send requests to one machine. For users with compatible hardware, that could provide a way to experiment with open models without paying for hosted inference or keeping a separate computer running. The project says requests are served on the local network, but users should not treat that as a guarantee of privacy or security without checking their own network setup and the app’s behavior.
Its significance is limited by the entry requirements. PS5LM is not a standard PlayStation app: the project requires a jailbroken console and specified homebrew payloads. It also does not establish that every GGUF model will perform well or fit in available memory. The reported speeds show what the project says it achieved on selected models, not a promise of equivalent performance across consoles, model sizes or settings.
PlayStation 5 jailbroken accessories
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Homebrew, llama.cpp and Console Hardware
GGUF is a model-file format used by llama.cpp, an open-source project for running language models locally. PS5LM combines that software with PS5-specific graphics support and a TV-oriented interface. Its source credits contributors to earlier PS5 homebrew and graphics work, including the native GPU app PS5LM Plus and the PS5_Vulkan and RADV efforts. The repository says it is licensed under GPL-3.0-or-later and is not affiliated with Sony Interactive Entertainment.
The project lists a supported firmware range of 7.00 to 13.60 and says the console must be jailbroken. Its setup instructions also call for payloads to support homebrew and install the app, with other tools recommended for transferring files and collecting hardware readings. Models can be downloaded through the app or copied onto console storage or a USB drive. This makes the release a specialized homebrew deployment, rather than an official feature available to ordinary PS5 owners.
““Your PS5 runs the language model. Every device on your network uses it.””
— PS5LM project description
As an affiliate, we earn on qualifying purchases.
Performance and Compatibility Questions
The source provides selected performance figures but does not detail the complete test conditions, repeatability, or performance across different PS5 models. It is also unclear how the app behaves under sustained workloads, how much performance varies with context length, or whether other GGUF architectures and quantizations will fit or run at useful speeds. The project says it supports architectures and quantizations known to llama.cpp, but that broad compatibility statement is not a guarantee for every model file.
The supplied information does not include an independent security review, a full account of network-access controls, or details on how the server handles access from devices beyond the intended local clients. The firmware and jailbreak requirements also mean availability depends on each user’s console configuration. The project’s reported benchmarks and privacy benefits should be understood as developer claims until separately tested.
As an affiliate, we earn on qualifying purchases.
Release Testing and Project Updates
Readers interested in the app can check the project’s latest release, installation notes and issue tracker for changes, supported models and reported problems. PS5LM’s instructions direct users to install the homebrew app, load a model, and connect a browser or compatible API client to the console’s local address. The project also points to documentation for using coding tools and to a roadmap and changelog.
No independent testing, formal product launch schedule or next release date is provided in the source material. Further confirmation would require hands-on tests across console models and GGUF files, alongside clearer documentation of benchmark conditions and network behavior. Until then, the reported performance and broad model support remain claims from the project itself.
As an affiliate, we earn on qualifying purchases.
Key Questions
What does PS5LM do?
PS5LM runs llama.cpp on a jailbroken PS5 and makes a loaded GGUF model available to other devices on the same network through compatible APIs or a browser chat page.
Can it run on any PS5?
No. The project specifies a jailbroken PS5 on firmware 7.00 to 13.60, along with required homebrew payloads. The source does not establish compatibility with every console configuration.
Does the console need an internet connection to answer prompts?
The project describes local inference that needs no cloud service while running. Model downloads or other setup tasks may involve network access, and the source does not describe every offline-use condition.
How fast are the models?
PS5LM reports 55 tokens per second for Qwen 3.5 9B at Q4_K_M and about 20 tokens per second for Qwen 3.8 27B at UD-Q2_K_XL on a PS5 Slim. These are project-reported results; test conditions and independent verification are not provided.
Is PS5LM an official Sony app?
No. The repository says the project is not affiliated with Sony Interactive Entertainment and presents itself as homebrew software.
Source: hn
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
