How To · 1
SoloRoleplayer// works with the model you run. The tested, recommended one is Mistral Nemo, abliterated.
Model
natong19/Mistral-Nemo-Instruct-2407-abliteratedThat page is the full-precision model (about 25 GB). For a normal computer, search Hugging Face for a 4-bit build of the same model in the format your server uses (GGUF for llama.cpp, KoboldCPP, LM Studio and Ollama; MLX for Apple silicon). A 4-bit build is about 7 to 8 GB.
A model whose refusal behavior has been removed, so it does not stop a story to lecture. It also means nothing filters its output: you decide how you use it, and SoloRoleplayer// is for adults, 18 and over.
Plan on 16 GB of RAM or more for a 12B model in 4-bit. Close heavy apps while you play. On 16 GB of RAM the model may run RAM hot and stop inside of SoloRoleplayer even if the LLM server keeps running. You do not have to exit your session, just restart the LLM server (Ollama, Cute LM, etc.) and continue from the error once it's live again. It's a known limitation of smaller RAM allotments but from my testing (On a Mac) has been safe to run in multi-hour sessions. On average 2 restarts for every few hours of play.
Yes, any model your server can run. The app, the companion setup and Anti-H 2.0 are tuned for Mistral Nemo, so results vary elsewhere.
A solo project. SoloRoleplayer// is for adults, 18 and over. It runs offline on your own computer and is not monitored: you are responsible for how you use it.