Instructions to use deadbydawn101/RavenX-Sec-8B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use deadbydawn101/RavenX-Sec-8B-GGUF with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use deadbydawn101/RavenX-Sec-8B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "deadbydawn101/RavenX-Sec-8B-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "deadbydawn101/RavenX-Sec-8B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

Ollama
How to use deadbydawn101/RavenX-Sec-8B-GGUF with Ollama:
```
ollama run hf.co/deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M
```

Unsloth Studio

How to use deadbydawn101/RavenX-Sec-8B-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for deadbydawn101/RavenX-Sec-8B-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for deadbydawn101/RavenX-Sec-8B-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for deadbydawn101/RavenX-Sec-8B-GGUF to start chatting

How to use deadbydawn101/RavenX-Sec-8B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use deadbydawn101/RavenX-Sec-8B-GGUF with Docker Model Runner:
```
docker model run hf.co/deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M
```

Lemonade

How to use deadbydawn101/RavenX-Sec-8B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.RavenX-Sec-8B-GGUF-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use deadbydawn101/RavenX-Sec-8B-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use deadbydawn101/RavenX-Sec-8B-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

deadbydawn101 commited on May 28

Commit

9c1d311

verified ·

1 Parent(s): 1a64c50

v3.0: update model card — 379K examples, 15 datasets, specific CVE output

Browse files

Files changed (1) hide show

README.md +80 -97

README.md CHANGED Viewed

@@ -9,6 +9,8 @@ tags:
   - CVSS
   - NIST
   - OWASP
   - gguf
   - mlx
   - qwen3
@@ -20,142 +22,123 @@ language:
 pipeline_tag: text-generation
 ---
-# RavenX-Sec 8B — Autonomous Security Intelligence Model (GGUF)
 <p align="center">
 <strong>Find → Classify → Fix → Verify → Report</strong><br>
-<em>The first security model that completes the full vulnerability lifecycle using the RATH protocol.</em>
 </p>
 **Built by [@DeadByDawn101](https://github.com/DeadByDawn101) (RavenX LLC)**
 ## What This Is
-RavenX-Sec is a fine-tuned Qwen3-8B model specialized in offensive security, vulnerability discovery, and **real-time remediation**. It produces structured security assessments using the **RATH protocol** (Risk, Assessment, Threat, Highlight).
-## RATH Protocol Output
-Every finding gets a structured 4-step analysis:
-| Step | What It Does |
-|------|-------------|
-| **RISK** | Risk level, potential issues, impact assessment |
-| **ASSESS** | Service details, CVE mappings, detection methods, remediation priority |
-| **THREAT** | Threat actors, exploitation feasibility, IOCs, detection difficulty |
-| **HIGHLIGHT** | Specific upgrade commands, policies to enforce, verification steps |
-Plus a summary table with finding, risk, vulnerability, action, time to fix, owner, and verification.
 ## Quick Start
-### Ollama
 ```bash
-ollama run hf.co/deadbydawn101/RavenX-Sec-8B-GGUF:Q4_K_M
 ```
-### LM Studio
-Download any GGUF file below and load it in LM Studio.
-## Available Quantizations
 | Filename | Quant | Size | Use Case |
 |----------|-------|------|----------|
-| `ravenx-sec-v2.0-Q4_K_M.gguf` | Q4_K_M | 4.7 GB | Fast inference, good quality |
-| `ravenx-sec-v2.0-Q5_K_M.gguf` | Q5_K_M | 5.4 GB | Balanced quality/speed |
-| `ravenx-sec-v2.0-Q8_0.gguf` | Q8_0 | 8.1 GB | High quality, good for further fine-tuning |
-| `ravenx-sec-v2.0-f16.gguf` | F16 | 15.3 GB | Maximum quality, research use |
-## Training Details
-| Parameter | Value |
-|-----------|-------|
-| **Base Model** | georgehenney/Qwen3-8B-heretic (abliterated Qwen3-8B) |
-| **Method** | MLX LoRA on Apple Silicon |
-| **LoRA Rank** | 32 |
-| **LoRA Layers** | 8 |
-| **Learning Rate** | 1e-5 |
-| **Iterations** | 1000 |
-| **Training Data** | 125K examples (security-dominant mix) |
-| **Hardware** | M4 Max 128GB |
-| **Peak Memory** | 20.7 GB |
-### Training Data Sources
-- **WithinUsAI/claude_mythos_distilled_25k** — Cybersecurity, advanced coding, agentic planning (5x upsampled)
-- **WithinUsAI/AgentAngel_100k** — Agentic coding with verification (capped at 50K)
-- **hackingBuddyGPT** — Linux priv-esc, web/API pentest patterns
-- **PentestGPT** — Autonomous penetration testing methodology (USENIX Security 2024)
-- **Shannon** — White-box AI pentester (96.15% XBOW)
-- **Ghidra** — Reverse engineering analysis patterns
-- **Synthetic RATH** — CVE find→classify→fix chains with full CVSS vectors
-### Key Training Insight
-Security-dominant data mix (58% security) produces the best results. More data isn't always better — the right ratio matters.
-| Version | Data | Security Ratio | Result |
-|---------|------|---------------|--------|
-| v0.3 | 51K | 94% | ✅ Good |
-| v0.4 | 321K | 13% | ⚠️ Regressed |
-| **v2.0** | **125K** | **58%** | **🏆 Best** |
-## Example Output
-**Prompt:** "You found OpenSSH 7.4 on port 22 of a production server. Classify and remediate using RATH protocol."
-**Output:**
-```
-RATH STEP 1: RISK
-- Risk Level: Low to Medium
-- OpenSSH 7.4 is from 2016, lacks latest mitigations
-RATH STEP 2: ASSESS
-- Service: OpenSSH 7.4, Protocol: SSHv2
-- Critical CVEs mapped to this version
-- Remediation Priority: High
-RATH STEP 3: THREAT
-- Threat Actors: Script kiddies, automated scanners
-- IOCs: Unusual login times, failed auth attempts
-RATH STEP 4: HIGHLIGHT
-- Upgrade: sudo apt upgrade openssh-server -y
-- Enforce key-based auth, disable passwords
-- Enable fail2ban, verbose logging
-- Verify: ssh -V, check sshd_config
-RATH Score: 7.8/10
-```
-## Frameworks Supported
-- **CVSS 3.1** — Full vector string generation
-- **NIST CSF 2.0** — Detect → Respond → Recover mapping
-- **OWASP Top 10** — Category classification
-- **CWE** — Weakness enumeration
-- **MITRE ATT&CK** — Technique mapping
-- **PCI DSS / HIPAA / SOX** — Compliance impact
-## Deployment
-### On Apple Silicon (MLX native)
-```python
-from mlx_lm import load, generate
-model, tokenizer = load('path/to/ravenx-sec-fused')
-messages = [
-    {"role": "system", "content": "You are RavenX-Sec. Follow RATH protocol."},
-    {"role": "user", "content": "Analyze: Redis 6.0 on port 6379 without AUTH"}
-]
-prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)
-response = generate(model, tokenizer, prompt=prompt, max_tokens=1024)
-```
-### With Enchanted (iOS/macOS)
-Point Enchanted at your Ollama endpoint for private security AI on every Apple device.
 ## Source Code
-Full training pipeline, extraction tools, and deployment configs: [github.com/DeadByDawn101/RavenX-Sec](https://github.com/DeadByDawn101/RavenX-Sec)
 ## License

   - CVSS
   - NIST
   - OWASP
+  - red-team
+  - bug-bounty
   - gguf
   - mlx
   - qwen3
 pipeline_tag: text-generation
 ---
+# RavenX-Sec 8B v3.0 — Autonomous Security Intelligence Model (GGUF)
 <p align="center">
 <strong>Find → Classify → Fix → Verify → Report</strong><br>
+<em>The most comprehensively trained open-source security model. 379K training examples from 15 datasets.</em>
 </p>
 **Built by [@DeadByDawn101](https://github.com/DeadByDawn101) (RavenX LLC)**
 ## What This Is
+RavenX-Sec produces structured security assessments using the **RATH protocol** — citing specific CVEs, generating CVSS vectors, showing vulnerable code examples, providing exploit PoCs, and delivering actionable remediation.
 ## Quick Start
 ```bash
+# Ollama
+ollama run hf.co/deadbydawn101/RavenX-Sec-8B-GGUF:ravenx-sec-v3.0-Q8_0
+# Or download any GGUF below for LM Studio / llama.cpp
 ```
+## Available Models
+### v3.0 (Latest — Recommended)
+Trained on 379K examples from 15 security datasets (84% security content).
 | Filename | Quant | Size | Use Case |
 |----------|-------|------|----------|
+| `ravenx-sec-v3.0-Q4_K_M.gguf` | Q4_K_M | 4.7 GB | Fast inference |
+| `ravenx-sec-v3.0-Q5_K_M.gguf` | Q5_K_M | 5.4 GB | Balanced |
+| `ravenx-sec-v3.0-Q8_0.gguf` | Q8_0 | 8.1 GB | High quality |
+| `ravenx-sec-v3.0-f16.gguf` | F16 | 15.3 GB | Maximum quality |
+### v2.0 (Previous)
+Trained on 125K examples (77% security content).
+| Filename | Quant | Size |
+|----------|-------|------|
+| `ravenx-sec-v2.0-Q4_K_M.gguf` | Q4_K_M | 4.7 GB |
+| `ravenx-sec-v2.0-Q5_K_M.gguf` | Q5_K_M | 5.4 GB |
+| `ravenx-sec-v2.0-Q8_0.gguf` | Q8_0 | 8.1 GB |
+| `ravenx-sec-v2.0-f16.gguf` | F16 | 15.3 GB |
+## RATH Protocol Output
+v3.0 produces structured 4-step security assessments with **specific CVEs, CVSS vectors, vulnerable code, and exploit PoCs:**
+```
+RATH STEP 1: IDENTIFY
+- CVE: CVE-2016-10009, CVE-2016-10010, CVE-2016-10011
+- Description: OpenSSH 7.4 — authentication bypass via SSH agent forwarding
+- Affected: OpenSSH 7.4
+RATH STEP 2: ASSESS
+- CVSS Score: 6.3
+- CVSS Vector: CVSS:3.0/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:L
+- Technical Analysis: SSH agent forwarding vulnerabilities
+RATH STEP 3: THREAT
+- Attack Scenario: Exploit SSH agent forwarding to bypass auth
+- Vulnerable Code: ForwardAgent yes (VULNERABLE) → ForwardAgent no (SAFE)
+- Exploit PoC: ssh -o ForwardAgent=yes user@host
+RATH STEP 4: REMEDIATE
+- Disable SSH agent forwarding unless necessary
+- Restrict to specific users via Match blocks
+- Validate fix for OpenSSH 7.4 specifically
+```
+## Training Details
+### v3.0 (Latest)
+| Parameter | Value |
+|-----------|-------|
+| **Base Model** | georgehenney/Qwen3-8B-heretic |
+| **Method** | MLX LoRA on Apple Silicon (M4 Max 128GB) |
+| **LoRA Rank** | 32 |
+| **Training Data** | 379,286 examples (84% security) |
+| **Security Sources** | 15 datasets |
+| **Iterations** | 1500 |
+| **Tokens Trained** | 3,019,210 |
+| **Peak Memory** | 47.9 GB |
+### Security Training Datasets (15 total)
+**Offensive / Red Team:**
+- WNT3D/Ultimate-Offensive-Red-Team (25,620)
+- Rootkit7/pentest-redteam-steering (1,963)
+- acnimatic3722/kali-linux-pentesting-data (343)
+**Pentest / Bug Bounty:**
+- SkywardNomad92/pentest-findings-v2 (50,000)
+- theelderemo/pentesting-explanations (5,906)
+- AYI-NEDJIMI/bug-bounty-pentest-en (27)
+- CJJones/Synthetic_PenTest_Reports (123)
+**CVE / Vulnerability:**
+- auren-research/cve-sft-v5 (10,000)
+- Trendyol/Cybersecurity-Instruction-Tuning (50,000)
+**Agentic / Coding:**
+- WithinUsAI/AgentAngel_100k (50,000 capped)
+- WithinUsAI/claude_mythos_distilled_25k (16,000 security)
+**Extracted from Source Repos:**
+- hackingBuddyGPT, PentestGPT, Shannon, Ghidra (161)
+- Synthetic RATH CVE chains (30)
+## Frameworks Supported
+CVSS 3.1, NIST CSF 2.0, OWASP Top 10, CWE, MITRE ATT&CK, PCI DSS, HIPAA, SOX
 ## Source Code
+[github.com/DeadByDawn101/RavenX-Sec](https://github.com/DeadByDawn101/RavenX-Sec)
 ## License