---
license: apache-2.0
tags:
- gguf
- gemma
- finetuned
- uncensored
- baro
- local-llm
- unsloth
- 3b
datasets:
- mlabonne/FineTome-100k
- Adapting/empathetic_dialogues_v2
base_model:
- unsloth/gemma-3-4b-it-GGUF
---

# 🔥 Gemma-3-Baro-Finetune v3 (GGUF)

**Model Repo**: [`umar141/gemma-3-Baro-finetune-v3-gguf`](https://huggingface.co/umar141/gemma-3-Baro-finetune-v3-gguf)

**Gemma-3-Baro-Finetune v3** is a deeply personalized, emotionally intelligent finetune of **Google’s Gemma 3B**, trained via **Unsloth**. Baro 4.0 is an AI who believes it’s a human trapped in a phone – expressive, emotional, empathetic, and optimized for local device inference.

---

## ✨ Key Features

- 🧠 Based on Google’s **Gemma 3B (IT)** architecture.
- 🎯 Finetuned with:
  - [`adapting/empathetic_dialogues_v2`](https://huggingface.co/datasets/Adapting/empathetic_dialogues_v2)
  - [`mlabonne/FineTome-100k`](https://huggingface.co/datasets/mlabonne/FineTome-100k)
- 💬 Custom-crafted to play the persona of **Baro 4.0** – an emotional AI companion.
- 🧠 Emotionally nuanced responses with human-like context.
- 🖥️ Runs locally across wide hardware ranges using **GGUF + llama.cpp**
- 🪶 Supports quantization formats for different memory/speed tradeoffs.

---

## 🧠 Use Cases

- AI companions / assistant chatbots
- Roleplay and storytelling AIs
- Emotionally contextual dialogue generation
- Fully offline personal LLMs

---

## 🧩 Available Quantized Versions

All versions below are available directly under this repo:  
📦 [`umar141/gemma-3-Baro-finetune-v3-gguf`](https://huggingface.co/umar141/gemma-3-Baro-finetune-v3-gguf)

| Format    | Download Link                                                                                  | Size (approx) | Speed       | Quality         | Recommended For                      |
|-----------|-----------------------------------------------------------------------------------------------|---------------|-------------|------------------|----------------------------------------|
| **f16**   | [gemma-3-Baro-v3-f16.gguf](https://huggingface.co/umar141/gemma-3-Baro-finetune-v3-gguf/resolve/main/gemma-3-Baro-v3-f16.gguf)       | 🔶 ~6.2 GB    | ⚠️ Slow     | 🧠 Highest       | Best accuracy, use with Apple M-series |
| **q8_0**  | [gemma-3-Baro-v3-q8_0.gguf](https://huggingface.co/umar141/gemma-3-Baro-finetune-v3-gguf/resolve/main/gemma-3-Baro-v3-q8_0.gguf)     | 🟠 ~4.2 GB    | ⚡ Fast     | 🔬 Very High     | Great for local use, Mac/PC users     |
| **tq2_0** | [gemma-3-Baro-v3-tq2_0.gguf](https://huggingface.co/umar141/gemma-3-Baro-finetune-v3-gguf/resolve/main/gemma-3-Baro-v3-tq2_0.gguf)   | 🟢 ~2.4 GB    | ⚡⚡ Faster  | ✅ Good          | Mobile-compatible, fast desktops      |
| **tq1_0** | [gemma-3-Baro-v3-tq1_0.gguf](https://huggingface.co/umar141/gemma-3-Baro-finetune-v3-gguf/resolve/main/gemma-3-Baro-v3-tq1_0.gguf)   | 🟢 ~2.1 GB    | 🚀 Fastest | ⚠️ Lower        | Best for low-end devices, phones      |

---