richardyoung/ mythos-qwen-1.5b-abliterated:Q4_K_M

210 Downloads Updated 1 month ago

Abliterated Mythos Engine 1.5B: a Qwen2.5-Coder security fine-tune (CVE, CTF, vuln research) decensored with Heretic. 2/100 refusals, KL 0.023.

tools

ollama run richardyoung/mythos-qwen-1.5b-abliterated:Q4_K_M

curl http://localhost:11434/api/chat \
  -d '{
    "model": "richardyoung/mythos-qwen-1.5b-abliterated:Q4_K_M",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='richardyoung/mythos-qwen-1.5b-abliterated:Q4_K_M',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'richardyoung/mythos-qwen-1.5b-abliterated:Q4_K_M',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 1 month ago

1 month ago

3ef24648b28e · 986MB

model

archqwen2

·

parameters1.54B

·

quantizationQ4_K_M

986MB

template

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

1.2kB

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

Readme

mythos-qwen-1.5b-abliterated

Abliterated Mythos Engine 1.5B: a Qwen2.5-Coder security fine-tune (CVE, CTF, vuln research) decensored with Heretic. ²⁄₁₀₀ refusals, KL 0.023.

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-qwen-1.5b-abliterated

Tag	Quant	Size	Context
`latest`, `Q4_K_M`	Q4_K_M	986 MB	32K native (8K default; raise with `/set parameter num_ctx`)

About

Mythos Engine is a Qwen2.5-Coder-1.5B-Instruct fine-tune for security work: CVE analysis, CTF, code security and vulnerability research.

Architecture: Qwen2 · 1.5B
Original model: expper/mythos-qwen-1.5b-final
Full weights: richardyoung/mythos-qwen-1.5b-final-heretic
MLX: richardyoung/mythos-qwen-1.5b-abliterated-mlx

Best for: Security research, CTF practice, code review for vulnerabilities, on very small hardware.

Abliteration

Made with Heretic v1.4.0.

Metric	This model	Original
Refusals	²⁄₁₀₀	⁹³⁄₁₀₀
KL divergence	0.0234	0

License

apache-2.0, inherited from the original model.

Built & maintained by Richard Young · DeepNeuro

# mythos-qwen-1.5b-abliterated

Abliterated Mythos Engine 1.5B: a Qwen2.5-Coder security fine-tune (CVE, CTF, vuln research) decensored with Heretic. 2/100 refusals, KL 0.023.

> **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-qwen-1.5b-abliterated
```

| Tag | Quant | Size | Context |
| :-- | :-- | :-- | :-- |
| `latest`, `Q4_K_M` | Q4_K_M | 986 MB | 32K native (8K default; raise with `/set parameter num_ctx`) |

## About

Mythos Engine is a Qwen2.5-Coder-1.5B-Instruct fine-tune for security work: CVE analysis, CTF, code security and vulnerability research.

**Architecture:** Qwen2 · 1.5B  
**Original model:** [expper/mythos-qwen-1.5b-final](https://huggingface.co/expper/mythos-qwen-1.5b-final)  
**Full weights:** [richardyoung/mythos-qwen-1.5b-final-heretic](https://huggingface.co/richardyoung/mythos-qwen-1.5b-final-heretic)  
**MLX:** [richardyoung/mythos-qwen-1.5b-abliterated-mlx](https://huggingface.co/richardyoung/mythos-qwen-1.5b-abliterated-mlx)

**Best for:** Security research, CTF practice, code review for vulnerabilities, on very small hardware.

## Abliteration

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

| Metric | This model | Original |
| :-- | :--: | :--: |
| Refusals | 2/100 | 93/100 |
| KL divergence | 0.0234 | 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)