README.md

---
license: mit
base_model: Qwen/Qwen2.5-Coder-1.5B
tags:
  - cybersecurity
  - offensive-security
  - code-generation
  - qwen2.5
  - gguf
  - ollama
---

# mythos-coder — Q4_K_M (GGUF)

Fine-tuned Qwen2.5-Coder-1.5B that writes Python exploit-detection code for the
MYTHOS autonomous offensive security platform.

- **Base model**: Qwen/Qwen2.5-Coder-1.5B
- **Quantization**: Q4_K_M
- **Size**: 941 MB (986,048,064 bytes)
- **Context**: 32,768
- **SHA-256**: `58e8bbd6c22aac2fbb0daf921b1329d75d72a4a397197b01f19f039363b656a5`

## System prompt

```
You are MYTHOS, an autonomous offensive security AI that writes Python code to
detect and exploit vulnerabilities. You output ONLY valid Python code.
No explanations, no thinking tags.
```

## Usage with Ollama

```bash
ollama create mythos-coder -f Modelfile
# Modelfile:
#   FROM mythos-coder-Q4_K_M.gguf
#   PARAMETER temperature 0.1
#   PARAMETER top_p 0.95
#   PARAMETER num_ctx 2048
```

## Usage in MYTHOS

The MYTHOS project (`scripts/provision_models.sh`) consumes this file
automatically via the URL below — set `MYTHOS_GGUF_URL` and the model builds
itself during setup:

```
MYTHOS_GGUF_URL=https://huggingface.co/zero2root/mythos-coder/resolve/main/mythos-coder-Q4_K_M.gguf
```

## Ethical use

For authorized penetration testing and defensive research only. Use only
against systems you own or are explicitly permitted to test.