Gemma-3-27B-Character-Creator-V2-GGUF/README.md

---
base_model:
- google/gemma-3-27b-it
tags:
- text-generation-inference
- transformers
- unsloth
- mistral
- roleplay
- sillytavern
- characters
license: gemma
language:
- en
---

This is a model made to create characters that can be used in Sillytavern, cai, jai and other such roleplay scenarios. The resulting characters should be about ~2k tokens and follow a prebaked structure.

Versions:
- 8B llama 3.3 based and [GGUFs](https://huggingface.co/SufficientPrune3897/Llama-3.3-8B-Character-Creator-V2-GGUF)
- 12B gemma 3 based and [GGUFs](https://huggingface.co/SufficientPrune3897/Gemma-3-12B-Character-Creator-V2-GGUF)
- 24B mistral small 3.2 based and GGUFs
- (this one) 27B gemma 3 based and GGUFs

How to use it:
- Simply tell the model what you want your character to be.
- It should know many popular franchises, the bigger the model, the more it knows.
- Fully uncensored.
- Asking for a different structure than the one the model uses might significantly reduce result quality.
- While follow up questions are supported, you will often get better results adjusting your original prompt.
- Supports asking for: prompts for pictures of the char, asking for changes and making an intro.

Changes [from V1](https://huggingface.co/SufficientPrune3897/gemma-3-27b-Character-Creator)
- No longer supports Groups and scenarios
- Characters should be much better
- It actually follows a structure and doesnt start making shit up after ~1k tokens

V3 and beyond:
The next version will either reintroduce scenarios, groups or feature reasoning. Probably both. Perhaps even lorebooks, although I'm still unsure how to execute on that... After that I will probably make my own real roleplay finetune or something.

If anybody wants support of their native language just ask me and tell me what model does the best for that.

I am very much open for feedback. A single comment can easily change how I will do my next version.


# Uploaded finetuned model

- **Developed by:** SufficientPrune3897
- **License:** apache-2.0
- **Finetuned from model :** google/gemma-3-27b-it

This mistral model was trained 2x faster with [Unsloth](https://github.com/unslothai/unsloth) and Huggingface's TRL library.

[<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200"/>](https://github.com/unslothai/unsloth)