# AI use cases for PS5 LM (cobanov/PS5LM) Repo: [cobanov/PS5LM](https://github.com/cobanov/PS5LM) | Use cases: 4 | Page: https://dedupedai.com/use-cases/cobanov-ps5lm/ PS5LM is an open-source llama.cpp app that runs GGUF language and decision models on a jailbroken PS5's GPU and serves them over the local network. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Run a private model server on your PS5: https://dedupedai.com/use-cases/cobanov-ps5lm/run-private-model-server-on-ps5/ - Point Claude Code at your PS5 model: https://dedupedai.com/use-cases/cobanov-ps5lm/use-ps5-model-with-claude-code/ - Classify support messages with a decision model: https://dedupedai.com/use-cases/cobanov-ps5lm/classify-support-messages-with-decision-model/ - Find the largest model your PS5 can run: https://dedupedai.com/use-cases/cobanov-ps5lm/benchmark-models-that-fit-ps5/ ## How to run a private model server on your PS5 with PS5 LM Repo: [cobanov/PS5LM](https://github.com/cobanov/PS5LM) | Difficulty: Advanced | Time: 3 hours | Basis: From the README Load a GGUF model on the PS5 GPU and use it from any device on your home network. **Who it is for:** A home user with a jailbroken PS5 who wants a private AI model for laptops and phones without a PC or cloud account. **The problem:** Cloud model services need an account and send your prompts off the network, and a PC-based server has to stay on. **The outcome:** A GGUF model answers OpenAI- or Anthropic-style requests and the browser chat page from any device on the network, with no PC or account needed while it runs. ### Steps 1. Confirm the PS5 is on firmware 7.00 to 13.60 and jailbroken, with kstuff-lite, etaHEN and ShadowMountPlus loaded. 2. Download `ps5lm-app-vX.Y.Z.zip` from the latest release and unzip it. 3. Copy the `PPSA99575` folder to `/data/homebrew` over FTP; ShadowMountPlus then adds PS5 LM to the home screen. 4. Open PS5 LM, go to Models > Suggested, pick a model, and load it; wait until Home says Ready. 5. Open `http://:8081` in a browser, or point your OpenAI or Anthropic client at that address. ### Prompt to paste into your agent ```text Using the cobanov/PS5LM repo, help the user set up a model server on a PS5 at . Send a chat request to http://:8081/v1/chat/completions with any model name and confirm a non-empty reply comes back. ``` ### Caveat PS5 LM needs a jailbroken console on firmware 7.00 to 13.60, so it will not run on a stock PS5. Risk, reverse engineering: The setup requires a jailbroken PS5 and kernel-level homebrew payloads, which Sony does not support and which can void warranties or change console behavior. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/cobanov-ps5lm/run-private-model-server-on-ps5/ ## How to point Claude Code at your PS5 model with PS5 LM Repo: [cobanov/PS5LM](https://github.com/cobanov/PS5LM) | Difficulty: Advanced | Time: 3 hours | Basis: From the README Route Claude Code requests to the PS5 model server through its Anthropic-compatible endpoint. **Who it is for:** A developer who wants to use Claude Code with a local model on a jailbroken PS5 instead of a cloud API. **The problem:** Coding agents normally send code to a hosted API, and the developer wants requests to stay on their own hardware. **The outcome:** Claude Code sends its requests to the PS5 at /v1/messages and receives replies from the loaded model. ### Steps 1. Set up the PS5 model server and load a model, as in the private model server use case, until Home says Ready. 2. Read docs/CLI.md in the repo for how coding tools such as opencode, Codex and Claude Code connect. 3. Point Claude Code at the PS5 by setting its Anthropic base address to `http://:8081`. 4. Send a short coding question from Claude Code and check that the reply appears on the PS5 TV as a chat. ### Prompt to paste into your agent ```text Using the cobanov/PS5LM repo, configure Claude Code to use the PS5 at on port 8081 as its Anthropic endpoint. Expected output is a reply from the loaded model, and verify that the request appears in the chat on the TV. ``` ### Caveat The README states Claude Code works against the Anthropic endpoint, but the exact client settings are in docs/CLI.md, which this summary does not reproduce. Risk, reverse engineering: This depends on a jailbroken PS5 running homebrew payloads, which is outside Sony's supported configuration. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/cobanov-ps5lm/use-ps5-model-with-claude-code/ ## How to classify support messages with a decision model with PS5 LM Repo: [cobanov/PS5LM](https://github.com/cobanov/PS5LM) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README Send a message to the PS5 /v1/systemone API to get typed intent and yes-or-no answers. **Who it is for:** A support or operations person who wants fast, local intent and tone labels for incoming customer messages. **The problem:** Sorting customer messages by intent and mood by hand is slow, and sending them to a cloud model may not be acceptable. **The outcome:** The decision model returns a typed choice for intent and a yes-or-no answer for anger, in a fraction of a second, from the PS5 on your network. ### Steps 1. Load d1 Omni 600M from the Suggested list in Models on the PS5. 2. Wait for Home to say Ready, then send the request below with your own state and questions, replacing http:// with the PS5 address and port 8081. 3. Read the typed decision and probabilities in the response, and check that the intent matches the message. 4. Send the same message to the TV's chat view to confirm the request was logged on the console. ### Prompt to paste into your agent ```text Using the cobanov/PS5LM repo, send a POST to http://:8081/v1/systemone with a customer message as state and an intent choice question. Expected output is a typed decision with probabilities, and verify that the chosen intent is one of the criteria you supplied. ``` ### Caveat The decision endpoint is described in the README as matching Ollaya's API, and the README does not give accuracy figures for any model. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/cobanov-ps5lm/classify-support-messages-with-decision-model/ ## How to find the largest model your PS5 can run with PS5 LM Repo: [cobanov/PS5LM](https://github.com/cobanov/PS5LM) | Difficulty: Intermediate | Time: 2 hours | Basis: From the README Compare suggested models and watch console load to choose one that fits memory and speed. **Who it is for:** A hobbyist with a PS5 Slim who wants the best model that runs at a usable speed on the console. **The problem:** Model sizes and context lengths are hard to match to a console's memory, and guessing wastes downloads and time. **The outcome:** A model is chosen from the Suggested list using the measured token speeds, and the Console page confirms temperature and load during a run. ### Steps 1. Open Models > Suggested on the PS5 and compare the listed models with the README table of sizes and tokens per second. 2. Check the planner's context and cache choice for a candidate model, then download and load it. 3. Open the Console page and use L2 or R2 to view the last two minutes or the last hour while the model runs. 4. Note the temperature, power and GPU load readings for your console and keep the model that runs at a speed you accept. ### Prompt to paste into your agent ```text Using the cobanov/PS5LM repo, list the models in the Suggested section for a PS5 at and compare them with the README speed table. Expected output is one recommended model and its measured speed, and verify that it fits the free memory the planner reports. ``` ### Caveat The README speeds are measured on a PS5 Slim, so results on other console models may differ. Dual-use notice: only use this on software, sites and systems you own or are authorized to test. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/cobanov-ps5lm/benchmark-models-that-fit-ps5/