
PS5LM v0.9.0 turns PS5 into local AI server
PS5LM v0.9.0 runs llama.cpp models on a jailbroken PS5’s GPU and serves them across a home network through OpenAI- and Anthropic-compatible APIs. The release adds a redesigned controller-friendly TV interface, telemetry, image support, and decision models.
PS5LM is a compelling homebrew experiment, though its jailbreak requirement makes it a niche alternative to ordinary local-inference hardware.
- –GPU acceleration expands capacity to models such as Qwen 3.8 27B at roughly 20 tokens per second.
- –Network APIs let laptops, phones, browsers, and coding tools use the console as a private inference server.
- –The TV-native interface and telemetry make the PS5 feel like a dedicated appliance rather than a hacked command-line box.
- –Support is limited to jailbroken PS5 firmware 7.00–13.60, sharply restricting practical adoption.
- –Its strongest value is demonstrating how much dormant consumer hardware can be repurposed for local AI.
DISCOVERED
1h ago
2026-10-17
PUBLISHED
1h ago
2026-10-17
RELEVANCE
AUTHOR
MisterSnowCrash