PS5LM v0.11.0 lets PC GPUs join a PS5 model fleet and expands context limits
PS5LM v0.11.0 adds PC GPU fleet workers, raises model memory to 11.6 GiB, and stops jailbroken PS5s from sleeping while serving models.
PS5LM, a PlayStation 5 homebrew app that runs GGUF language models locally via llama.cpp on jailbroken consoles, released v0.11.0 on 2026-11-02. It targets jailbroken PS5 systems on firmware 7.00 to 13.60 using kstuff-lite, etaHEN, and ShadowMountPlus.
A PS5 can pair with a PC GPU as a fleet worker by building the host with PS5LM_FLEET_GPU=metal, cuda, or vulkan and running ggml-rpc-server. The fleet worker ships inside the app without a separate archive, controlled through /data/PS5LM/fleet-worker.txt. Memory allocation rises to a 12.0 GiB pool with 11.6 GiB planned for models, while maximum context reaches 192k and 256k. Sliding-window attention planning is fixed, and Bonsai 2 27B joins Suggested at a 64k context.
The release also prevents a serving PS5 from entering rest mode after an hour without controller input, which previously ended the app and left controllers unable to connect until a restart. To install, unzip ps5lm-app-v0.11.0.zip and copy PPSA99575 to /data/homebrew over FTP; OnionHEN users must add PPSA99575 to app_jailbreak.exact_title_ids in /data/OnionHEN/config.ini, though the OnionHEN fix has not yet been run on an OnionHEN console.