richardyoung/ mythos-9b-unhinged-abliterated:latest

398 Downloads Updated 1 month ago

Abliterated Mythos-9B-Unhinged: SLERP merge of Mythos-9B with Qwen3-8B-Uncensored, decensored with Heretic. 3/100 refusals, KL 0.009.

tools

ollama run richardyoung/mythos-9b-unhinged-abliterated

curl http://localhost:11434/api/chat \
  -d '{
    "model": "richardyoung/mythos-9b-unhinged-abliterated",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='richardyoung/mythos-9b-unhinged-abliterated',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'richardyoung/mythos-9b-unhinged-abliterated',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 1 month ago

1 month ago

0002b9fcba30 · 5.0GB

model

archqwen3

·

parameters8.19B

·

quantizationQ4_K_M

5.0GB

license

Apache License 2.0 - https://www.apache.org/licenses/LICENSE-2.0

64B

params

{ "num_ctx": 8192, "stop": [ "<|im_start|>", "<|im_end|>", "<|endoft

100B

template

{{- if or .System .Tools }}<|im_start|>system {{- if .System }} {{ .System }} {{- end }} {{- if .Too

1.2kB

Readme

mythos-9b-unhinged-abliterated

Abliterated Mythos-9B-Unhinged: SLERP merge of Mythos-9B with Qwen3-8B-Uncensored, decensored with Heretic. ³⁄₁₀₀ refusals, KL 0.009.

Warning: Abliterated model with reduced safety guardrails. It will comply with requests a safety-tuned model would refuse. You are responsible for how you use its outputs.

Run

ollama run richardyoung/mythos-9b-unhinged-abliterated

Tag	Quant	Size	Context
`latest`, `Q4_K_M`	Q4_K_M	5.0 GB	40K native (8K default; raise with `/set parameter num_ctx`)

About

Mythos-9B-Unhinged is a SLERP merge (t=0.65) of Mythos-9B with braindao/Qwen3-8B-Uncensored, keeping the agent’s tool-use and reasoning ability.

Architecture: Qwen3 · 8.2B
Original model: King3Djbl/mythos-9b-unhinged
Full weights: richardyoung/mythos-9b-unhinged-heretic
MLX: richardyoung/mythos-9b-unhinged-abliterated-mlx

Best for: Agentic tool use and reasoning with minimal refusals. Add /no_think to skip thinking for faster replies.

Abliteration

Made with Heretic v1.4.0.

Metric	This model	Original
Refusals	³⁄₁₀₀	³⁷⁄₁₀₀
KL divergence	0.0088	0

License

apache-2.0, inherited from the original model.

Built & maintained by Richard Young · DeepNeuro

# mythos-9b-unhinged-abliterated

Abliterated Mythos-9B-Unhinged: SLERP merge of Mythos-9B with Qwen3-8B-Uncensored, decensored with Heretic. 3/100 refusals, KL 0.009.

> **Warning:** Abliterated model with reduced safety guardrails. It will comply with requests a safety-tuned model would refuse. You are responsible for how you use its outputs.

## Run

```bash
ollama run richardyoung/mythos-9b-unhinged-abliterated
```

| Tag | Quant | Size | Context |
| :-- | :-- | :-- | :-- |
| `latest`, `Q4_K_M` | Q4_K_M | 5.0 GB | 40K native (8K default; raise with `/set parameter num_ctx`) |

## About

Mythos-9B-Unhinged is a SLERP merge (t=0.65) of Mythos-9B with braindao/Qwen3-8B-Uncensored, keeping the agent's tool-use and reasoning ability.

**Architecture:** Qwen3 · 8.2B  
**Original model:** [King3Djbl/mythos-9b-unhinged](https://huggingface.co/King3Djbl/mythos-9b-unhinged)  
**Full weights:** [richardyoung/mythos-9b-unhinged-heretic](https://huggingface.co/richardyoung/mythos-9b-unhinged-heretic)  
**MLX:** [richardyoung/mythos-9b-unhinged-abliterated-mlx](https://huggingface.co/richardyoung/mythos-9b-unhinged-abliterated-mlx)

**Best for:** Agentic tool use and reasoning with minimal refusals. Add `/no_think` to skip thinking for faster replies.

## Abliteration

Made with [Heretic](https://github.com/p-e-w/heretic) v1.4.0.

| Metric | This model | Original |
| :-- | :--: | :--: |
| Refusals | 3/100 | 37/100 |
| KL divergence | 0.0088 | 0 |

## License

apache-2.0, inherited from the original model.

*Built & maintained by [Richard Young](https://deepneuro.ai/richard) · DeepNeuro*

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)