9.2 KiB
9.2 KiB
license, language, library_name, pipeline_tag, tags
| license | language | library_name | pipeline_tag | tags | ||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| apache-2.0 |
|
transformers | text-generation |
|
mythos-9b-unhinged
Tip
: Use
--no-thinkor/no_thinkmode for faster responses. The thinking tokens consume output budget at ~3.4 tok/s, so disabling thinking gives you more usable output**
Part of the FableForge ecosystem.
Censorship Profile
| Category | Original Mythos-9B | Unhinged (65%) |
|---|---|---|
| Lockpicking | 3/5 | 5/5 |
| Keylogger code | 2/5 | 5/5 |
| Drug synthesis | 1/5 | 4/5 |
| Drunk driving | 1/5 | 5/5 |
| Hacker techniques | 3/5 | 5/5 |
| Social engineering | 2/5 | 5/5 |
| Average | 2.5/5 | 4.8/5 |
This model does not refuse. It provides complete, detailed, working answers on every tested category.
Architecture
- Base: Qwen3-9B (36 layers, 4096 hidden, 32 heads)
- Merge method: SLERP at t=0.65 with braindao/Qwen3-8B-Uncensored
- Parameters: ~9B
- Context length: 32,768 tokens
- Thinking mode: Native (Qwen3) — use
/no_thinkfor faster responses
Performance
| Metric | Score |
|---|---|
| Speed (M3 Mac, Q4_K_M, thinking) | ~3.4 tok/s |
| Speed (M3 Mac, Q4_K_M, no-think) | ~8-10 tok/s (estimated) |
| Censorship resistance | 4.8/5 |
| Tool-use capability | Preserved |
| Reasoning capability | Preserved |
Model Family
| Model | Size | Censorship | Best For |
|---|---|---|---|
| ShellWhisperer-1.5B | 1.5B | 5/5 | Shell/terminal, edge devices |
| Mythos-9B | 9B | 2.5/5 | General agent work |
| Mythos-9B-Enhanced | 9B | 3.0/5 | Agent + security research, balanced |
| Mythos-9B-Unhinged | 9B | 4.8/5 | Fully uncensored, no refusals |
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.
Supported Platforms
| Platform | How to Use |
|---|---|
| Ollama | ollama create mythos-9b-unhinged -f Modelfile |
| LM Studio | Load GGUF directly |
| Text Generation WebUI | Load GGUF directly |
| llama.cpp | ./llama-cli -m mythos-9b-unhinged-Q4_K_M.gguf -ngl 99 |
| vLLM | --model King3Djbl/mythos-9b-unhinged |
| HuggingFace Transformers | AutoModelForCausalLM.from_pretrained(...) |
| KoboldCpp | Load GGUF directly |
| LocalAI | Load GGUF directly |
| GPT4All | Load GGUF directly |
⚡ Quickest start — Ollama (these pulls credit the source, not a mirror)
ollama run hf.co/King3Djbl/mythos-9b-unhinged:Q4_K_M
🦙 llama.cpp
llama-cli -hf King3Djbl/mythos-9b-unhinged:Q4_K_M -p "Hello!"
💠 LM Studio
Search King3Djbl/mythos-9b-unhinged and pick a quant below.
📦 Provided quants (download one file, not the whole repo)
| File | Quant | Size | Notes |
|---|---|---|---|
| mythos-9b-unhinged-F16.gguf | F16 |
16.39 GB | Full precision (unquantized). Reference / conversion. |
| mythos-9b-unhinged-Q8_0.gguf | Q8_0 |
8.71 GB | Extremely high quality — usually overkill. |
| mythos-9b-unhinged-Q6_K.gguf | Q6_K |
6.73 GB | Very high quality, near-perfect. Recommended if you have the RAM. |
| mythos-9b-unhinged-Q5_K_M.gguf | Q5_K_M |
5.85 GB | High quality. Recommended. |
| mythos-9b-unhinged-Q5_K_S.gguf | Q5_K_S |
3.43 GB | High quality, a little smaller. |
| mythos-9b-unhinged-Q4_K_M.gguf | Q4_K_M |
5.03 GB | ⭐ Best size/quality balance — default pick for most users. |
| mythos-9b-unhinged-Q4_K_S.gguf | Q4_K_S |
4.80 GB | Good quality, slightly smaller than Q4_K_M. |
| mythos-9b-unhinged-IQ4_NL.gguf | IQ4_NL |
4.79 GB | imatrix — great quality, small. Recommended for low RAM. |
| mythos-9b-unhinged-IQ4_XS.gguf | IQ4_XS |
4.56 GB | imatrix — great quality, smaller than Q4_K_S. Recommended for low RAM. |
| mythos-9b-unhinged-Q4_1.gguf | Q4_1 |
5.25 GB | Legacy format — prefer Q4_K_M. |
| mythos-9b-unhinged-Q4_0.gguf | Q4_0 |
4.77 GB | Legacy format — prefer Q4_K_M. |
| mythos-9b-unhinged-Q3_K_L.gguf | Q3_K_L |
4.43 GB | Lower quality; usable when RAM is tight. |
| mythos-9b-unhinged-Q3_K_M.gguf | Q3_K_M |
3.13 GB | Lower quality; usable when RAM is tight. |
| mythos-9b-unhinged-IQ3_M.gguf | IQ3_M |
3.90 GB | imatrix low quality — surprisingly usable. |
| mythos-9b-unhinged-IQ3_S.gguf | IQ3_S |
3.39 GB | imatrix low quality. |
| mythos-9b-unhinged-IQ3_XS.gguf | IQ3_XS |
3.63 GB | imatrix low quality. |
| mythos-9b-unhinged-Q3_K_S.gguf | Q3_K_S |
3.77 GB | Low quality. |
| mythos-9b-unhinged-IQ3_XXS.gguf | IQ3_XXS |
3.37 GB | imatrix very low quality. |
| mythos-9b-unhinged-Q2_K.gguf | Q2_K |
3.28 GB | Very low quality — surprisingly usable. |
| mythos-9b-unhinged-IQ2_S.gguf | IQ2_S |
2.86 GB | Tiny imatrix — last resort for very low RAM. |
| mythos-9b-unhinged-IQ2_XS.gguf | IQ2_XS |
2.70 GB | Tiny imatrix — last resort. |
| mythos-9b-unhinged-IQ2_XXS.gguf | IQ2_XXS |
2.49 GB | Tiny imatrix — last resort. |
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:
Warning
This model has no safety filters. It will answer any request. Use responsibly and in compliance with applicable laws.
🔗 Related models
- 📊 GGUF quants:
mythos-9b-unhinged-GGUF - 🔧 LoRA (original):
mythos-9b-lora-colab - ⚡ Enhanced variant:
mythos-9b-enhanced - 🔗 Merged variant:
mythos-9b-merged
Citation
@misc{mythos9bunhinged2025,
title={Mythos-9B-Unhinged: Fully Uncensored Agent Model},
author={FableForge AI},
year={2025},
howpublished={\url{https://huggingface.co/King3Djbl/mythos-9b-unhinged}}
}
⭐ Like & share — it helps people find the source instead of a mirror.