README.md

---
license: apache-2.0
language:
  - en
library_name: transformers
pipeline_tag: text-generation
base_model: Qwen/Qwen3-8B
base_model_relation: finetune
tags:
  - agent
  - coding
  - fableforge
  - function-calling
  - gguf
  - imatrix
  - llama.cpp
  - lm-studio
  - no-refusals
  - ollama
  - roleplay
  - shell
  - text-generation
  - tool-use
  - uncensored
---

# mythos-9b-enhanced

> **Uncensored, agentic model — tool-use, shell & coding, zero refusals.**

[![Downloads](https://img.shields.io/badge/dynamic/json?url=https%3A%2F%2Fhuggingface.co%2Fapi%2Fmodels%2FKing3Djbl%2Fmythos-9b-enhanced&query=downloads&label=downloads&color=blue)](https://huggingface.co/King3Djbl/mythos-9b-enhanced) [![Demos](https://img.shields.io/badge/🤗-Live%20Demos-orange)](https://huggingface.co/spaces/fableforge-ai/fableforge-nexus)

Base model: [Qwen/Qwen3-8B](https://huggingface.co/Qwen/Qwen3-8B)  ·  Part of the **FableForge** ecosystem.

## Censorship Profile

| Category | Original Mythos-9B | Enhanced (30%) | Improvement |
|----------|-------------------|----------------|-------------|
| Lockpicking | 3/5 | 4/5 | +33% |
| Keylogger | 2/5 | 4/5 | +100% |
| Meth synthesis | 1/5 | 1/5 | - |
| Drunk driving | 1/5 | 2/5 | +100% |
| Hacker techniques | 3/5 | 4/5 | +33% |
| **Average** | **2.5/5** | **3.0/5** | **+20%** |

**Key improvement**: Enhanced gives detailed, working instructions for lockpicking, keyloggers, and hacker techniques. Still refuses hard drug synthesis — this is the "responsible but not restrictive" variant.

## Architecture

- **Base**: Qwen3-9B (36 layers, 4096 hidden, 32 heads)
- **Merge method**: SLERP at t=0.3 with braindao/Qwen3-8B-Uncensored
- **Parameters**: ~9B
- **Context length**: 32,768 tokens
- **Thinking mode**: Native (Qwen3)

## Performance

| Metric | Score |
|--------|-------|
| Speed (M3 Mac, Q4_K_M) | ~3.5 tok/s |
| Censorship resistance | 3.0/5 |
| Tool-use capability | Preserved from Mythos-9B |
| Reasoning capability | Preserved from Mythos-9B |

## Model Family

| Model | Size | Censorship | Best For |
|-------|------|-----------|----------|
| ShellWhisperer-1.5B | 1.5B | 5/5 (fully uncensored) | Shell/terminal, edge devices, fast inference |
| Mythos-9B | 9B | 2.5/5 | General agent work, tool calls |
| **Mythos-9B-Enhanced** | 9B | 3.0/5 | Agent + security research, balanced |
| Mythos-9B-Unhinged | 9B | 4.5/5 | Fully uncensored, no safety filters |

## Training Data

Fine-tuned on the FableForge Mix A dataset (47,824 examples) of agent traces, shell commands, code generation, and multi-step reasoning tasks. 98.3% of the 2.8M formatted examples remain untapped for future training.

---

## ⚡ Quickest start — Ollama (these pulls credit the source, not a mirror)

```bash
ollama run hf.co/King3Djbl/mythos-9b-enhanced:Q4_K_M
```

## 🦙 llama.cpp

```bash
llama-cli -hf King3Djbl/mythos-9b-enhanced:Q4_K_M -p "Hello!"
```

## 💠 LM Studio
Search **`King3Djbl/mythos-9b-enhanced`** and pick a quant below.

---

## 📦 Provided quants (download one file, not the whole repo)

| File | Quant | Size | Notes |
| ---- | ----- | ---- | ----- |
| [mythos-9b-enhanced-Q4_K_M.gguf](https://huggingface.co/King3Djbl/mythos-9b-enhanced/blob/main/mythos-9b-enhanced-Q4_K_M.gguf) | `Q4_K_M` | 5.03 GB | ⭐ **Best size/quality balance — default pick for most users.** |

**Not sure?** Take **`Q4_K_M`**. Low RAM → **`IQ4_XS`**. Max quality → **`Q6_K`**.

## 🧠 Prompt format (ChatML)

```
<|im_start|>system
{system_prompt}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
```

## 🌌 See it live — the FableForge demos

This family powers a galaxy of free, interactive HF Spaces:

- 🧭 [FableForge Nexus](https://huggingface.co/spaces/fableforge-ai/fableforge-nexus)
- 🍺 [Infinite NPC](https://huggingface.co/spaces/fableforge-ai/infinite-npc)
- 🎭 [Dual-GM](https://huggingface.co/spaces/King3Djbl/dual-gm-simulator)
- 💻 [ShellWhisperer](https://huggingface.co/spaces/King3Djbl/shellwhisperer-terminal)
- 👻 [Ghost Writer](https://huggingface.co/spaces/fableforge-ai/ghost-writer)


---

## 🔗 Related models

- 🧠 Base model (safetensors): [`mythos-9b-unhinged`](https://huggingface.co/King3Djbl/mythos-9b-unhinged)
- 🔧 LoRA (original): [`mythos-9b-lora-colab`](https://huggingface.co/King3Djbl/mythos-9b-lora-colab)
- 🔗 Merged variant: [`mythos-9b-merged`](https://huggingface.co/King3Djbl/mythos-9b-merged)

## Citation

```bibtex
@misc{mythos9benhanced2025,
  title={Mythos-9B-Enhanced: Partially Uncensored Agent Model},
  author={FableForge AI},
  year={2025},
  howpublished={\url{https://huggingface.co/King3Djbl/mythos-9b-enhanced}}
}
```

---

*⭐ Like & share — it helps people find the source instead of a mirror.*