Instructions to use deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx with MLX:

# Make sure mlx-lm is installed
# pip install --upgrade mlx-lm

# Generate text with mlx-lm
from mlx_lm import load, generate

model, tokenizer = load("deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx")

prompt = "Write a story about Einstein"
messages = [{"role": "user", "content": prompt}]
prompt = tokenizer.apply_chat_template(
    messages, add_generation_prompt=True
)

text = generate(model, tokenizer, prompt=prompt, verbose=True)

Notebooks
Google Colab
Kaggle
Local Apps Settings
LM Studio

Unsloth Studio

How to use deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx 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-35B-Security-RATH-mlx 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-35B-Security-RATH-mlx 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-35B-Security-RATH-mlx to start chatting

Load model with FastModel

pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name="deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx",
    max_seq_length=2048,
)

How to use deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx with Pi:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx"

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "mlx-lm": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx with Hermes Agent:

Start the MLX server

# Install MLX LM:
uv tool install mlx-lm
# Start a local OpenAI-compatible server:
mlx_lm.server --model "deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx"

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-35B-Security-RATH-mlx

Run Hermes

hermes

MLX LM

How to use deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx with MLX LM:

Generate or start a chat session

# Install MLX LM
uv tool install mlx-lm
# Interactive chat REPL
mlx_lm.chat --model "deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx"

Run an OpenAI-compatible server

# Install MLX LM
uv tool install mlx-lm
# Start the server
mlx_lm.server --model "deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx"
# Calling the OpenAI-compatible server with curl
curl -X POST "http://localhost:8000/v1/chat/completions" \
   -H "Content-Type: application/json" \
   --data '{
     "model": "deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx",
     "messages": [
       {"role": "user", "content": "Hello"}
     ]
   }'

deadbydawn101 commited on about 19 hours ago

Commit

be2f63f

verified ·

1 Parent(s): ae9a501

v5.0: README.md — 35B MoE, 680K, 6-step RATH

Browse files

Files changed (1) hide show

README.md +17 -219

README.md CHANGED Viewed

@@ -1,224 +1,22 @@
 ---
 license: apache-2.0
-base_model: huihui-ai/Huihui-Qwen3.6-35B-A3B-Claude-4.7-Opus-abliterated
-tags:
-  - security
-  - cybersecurity
-  - pentest
-  - CVSS
-  - OWASP
-  - red-team
-  - bug-bounty
-  - 128k-context
-  - MLX
-  - Safetensors
-  - apple-silicon
-  - ravenx
-  - rath-protocol
-  - vision
-  - MoE
-  - 35B
 language:
-  - en
-pipeline_tag: text-generation
 library_name: mlx
 ---
-# 🐦‍⬛ RavenX-Sec 35B v5.0 — Autonomous Security Intelligence (Vision + MoE)
-> **MLX** · Apple Silicon · 35B MoE (3B active) · Vision · 262K context · 6-step RATH protocol · 675,696 training examples · Claude 4.7 Opus reasoning distilled
-**The most powerful open-source security model.** Qwen3.6-35B-A3B with Claude Opus reasoning, abliterated for zero refusals, fine-tuned on 675,696 security examples from 21 HuggingFace datasets + 19 proprietary GitHub repos.
-> 🛡️ **Previous version:** [RavenX-Sec-8B v4.0](https://huggingface.co/deadbydawn101/RavenX-Sec-8B-Security-RATH-128k-mlx-4bit) (610K examples, 8B)
->
-> 💰 **Trading model:** [RavenX-Trade-8B v1.1](https://huggingface.co/deadbydawn101/RavenX-Trade-8B-MAP-128k-mlx-4bit) (318K examples, MAP protocol)
-**Built by [@DeadByDawn101](https://github.com/DeadByDawn101) · RavenX LLC**
----
-## What's New in v5.0
-| Feature | v4.0 (8B) | v5.0 (35B) |
-|---------|-----------|------------|
-| **Parameters** | 8B dense | **35B MoE (3B active)** |
-| **Vision** | Text only | **Screenshots, diagrams, code** |
-| **Base reasoning** | Qwen3-8B | **Claude 4.7 Opus distilled** |
-| **Context** | 128K | **262K native** |
-| **Uncensoring** | Heretic | **Abliterated (zero refusal)** |
-| **Training data** | 610K (21 HF datasets) | **675,696 (21 HF + 19 repos)** |
-| **Architecture** | Dense | **MoE — 35B brain, 3B speed** |
----
-## RATH Protocol
-| Step | What It Does |
-|------|-------------|
-| **R — Risk** | Identify vulnerability and attack surface |
-| **A — Assess** | CVSS scoring, CWE classification, severity rating |
-| **T — Threat** | MITRE ATT&CK mapping, exploit scenarios, threat actors |
-| **H — Highlight** | Remediation steps, code fixes, configuration changes |
-| **D — Document** | Executive summary, compliance mapping, SLA timelines |
-| **P — Prevent** | Monitoring rules, detection signatures, prevention controls |
----
-## Training Data — Complete Inventory (675,696 Examples)
-### HuggingFace Datasets (21 Sources — 610,220 Examples)
-#### Security Instruction Tuning (11 Datasets)
-| Dataset | Type |
-|---------|------|
-| [Trendyol/Cybersecurity-Instruction-Tuning-Dataset](https://huggingface.co/datasets/Trendyol/Trendyol-Cybersecurity-Instruction-Tuning-Dataset) | Enterprise cybersecurity instruction tuning |
-| [WNT3D/Ultimate-Offensive-Red-Team](https://huggingface.co/datasets/WNT3D/Ultimate-Offensive-Red-Team) | Offensive red team techniques and procedures |
-| [AYI-NEDJIMI/bug-bounty-pentest-en](https://huggingface.co/datasets/AYI-NEDJIMI/bug-bounty-pentest-en) | Bug bounty and pentesting data |
-| [auren-research/cve-sft-v5](https://huggingface.co/datasets/auren-research/cve-sft-v5) | CVE-specific SFT data |
-| [theelderemo/pentesting-explanations](https://huggingface.co/datasets/theelderemo/pentesting-explanations) | Pentesting technique walkthroughs |
-| [Rootkit7/pentest-redteam-steering](https://huggingface.co/datasets/Rootkit7/pentest-redteam-steering) | Red team steering and methodology |
-| [cpagac/venomx-pentesting-harmful](https://huggingface.co/datasets/cpagac/venomx-pentesting-harmful) | Advanced pentesting scenarios |
-| [SkywardNomad92/pentest-findings-v2](https://huggingface.co/datasets/SkywardNomad92/pentest-findings-v2) | Pentest findings with classification |
-| [acnimatic3722/kali-linux-pentesting-data](https://huggingface.co/datasets/acnimatic3722/kali-linux-pentesting-data) | Kali Linux tool usage and workflows |
-| [CJJones/Synthetic_PenTest_Reports](https://huggingface.co/datasets/CJJones/Synthetic_PenTest_Reports) | Synthetic penetration testing reports |
-| [nangyall/4-Security-Tools-Pentesting](https://huggingface.co/datasets/nangyall/4-Security-Tools-Pentesting) | Security tools usage (nmap, burp, etc) |
-#### Vulnerability & Code Security (4 Datasets)
-| Dataset | Type |
-|---------|------|
-| [ByteDance/PatchEval](https://huggingface.co/datasets/ByteDance/PatchEval) | 1000 real-world CVE patches, 65 CWE categories |
-| [bigcode/vuln-eval](https://huggingface.co/datasets/bigcode/vuln-eval) | Vulnerability recognition and repair |
-| [Fraser/cwe-benchmark](https://huggingface.co/datasets/Fraser/cwe-benchmark) | Source code mapped to CWE IDs |
-| [isek/cybersecurity-instructions](https://huggingface.co/datasets/isek/cybersecurity-instructions) | RE walkthroughs, security scripting |
-#### Agentic & Reasoning (3 Datasets)
-| Dataset | Type |
-|---------|------|
-| [WithinUsAI/claude_mythos_distilled_25k](https://huggingface.co/datasets/WithinUsAI/claude_mythos_distilled_25k) | Distilled reasoning (cybersecurity, agentic planning) |
-| [WithinUsAI/AgentAngel_100k](https://huggingface.co/datasets/WithinUsAI/AgentAngel_100k) | Agentic coding (plan, patch, verify, iterate) |
-| [HackerSignal/Threat-Intel](https://huggingface.co/datasets/HackerSignal/Threat-Intel) | Historical cybersecurity docs, exploit scripts |
-#### Infrastructure & Code (3 Datasets)
-| Dataset | Type |
-|---------|------|
-| [JetBrains-Research/commit-chronicle](https://huggingface.co/datasets/JetBrains-Research/commit-chronicle) | Git commits, security-tagged patches |
-| [bigcode/commitpack](https://huggingface.co/datasets/bigcode/commitpack) | Code changes across 300+ languages |
-| [bigcode/the-stack-v2](https://huggingface.co/datasets/bigcode/the-stack-v2) | Infrastructure-as-code (Ansible, Terraform, Shell) |
----
-### Proprietary GitHub Repos (19 Sources — 65,476 Examples)
-#### Autonomous Pentesting & Security (4 Repos)
-| Repo | Examples | Content |
-|------|----------|---------|
-| [DeadByDawn101/phalanx](https://github.com/DeadByDawn101/phalanx) | 65 | SWARM pentesting agents, RoE enforcement, MITRE ATT&CK |
-| [DeadByDawn101/chrome-devtools-mcp](https://github.com/DeadByDawn101/chrome-devtools-mcp) | 194 | Browser debugging, MCP protocol, security headers |
-| [DeadByDawn101/camofox-mcp](https://github.com/DeadByDawn101/camofox-mcp) | 134 | Anti-detection, browser fingerprint evasion |
-| [DeadByDawn101/phantom](https://github.com/DeadByDawn101/phantom) | 662 | Autonomous agent architecture, security model |
-#### Agent Frameworks (5 Repos)
-| Repo | Examples | Content |
-|------|----------|---------|
-| [nousresearch/hermes-agent](https://github.com/nousresearch/hermes-agent) | 42,929 | NousResearch self-improving agent patterns |
-| [kilo-org/kilocode](https://github.com/kilo-org/kilocode) | 3,224 | Agent framework, tool calling, code execution |
-| [Gitlawb/openclaude](https://github.com/Gitlawb/openclaude) | 310 | Coding agent patterns, task decomposition |
-| [DeadByDawn101/self-improving-agent](https://github.com/DeadByDawn101/self-improving-agent) | 131 | Agent learning loops, self-correction |
-| [DeadByDawn101/self_improving_coding_agent](https://github.com/DeadByDawn101/self_improving_coding_agent) | 743 | Code generation, testing, self-improvement |
-#### Research & Automation (4 Repos)
-| Repo | Examples | Content |
-|------|----------|---------|
-| [DeadByDawn101/AI-Scientist](https://github.com/DeadByDawn101/AI-Scientist) | 6,737 | Research automation, hypothesis generation |
-| [DeadByDawn101/AutoResearchClaw](https://github.com/DeadByDawn101/AutoResearchClaw) | 3,639 | Automated research pipelines |
-| [DeadByDawn101/autoresearch-mlx](https://github.com/DeadByDawn101/autoresearch-mlx) | 23 | MLX-native research tools |
-| [DeadByDawn101/get-shit-done-redux](https://github.com/DeadByDawn101/get-shit-done-redux) | 4,230 | Task management, agent orchestration |
-#### Performance & Optimization (4 Repos)
-| Repo | Examples | Content |
-|------|----------|---------|
-| [DeadByDawn101/turboquant-mlx](https://github.com/DeadByDawn101/turboquant-mlx) | 304 | KV cache compression, MLX optimization |
-| [DeadByDawn101/tokenspeed](https://github.com/DeadByDawn101/tokenspeed) | 1,950 | Token-level optimization, inference speed |
-| [DeadByDawn101/auto-antislop](https://github.com/DeadByDawn101/auto-antislop) | 78 | Output quality control at token level |
-| [DeadByDawn101/adhd](https://github.com/DeadByDawn101/adhd) | 95 | Attention management, focus optimization |
-#### Infrastructure (2 Repos)
-| Repo | Examples | Content |
-|------|----------|---------|
-| [DeadByDawn101/brane-code](https://github.com/DeadByDawn101/brane-code) | 11 | Distributed compute patterns |
-| [google-gemma/gemma-skills](https://github.com/google-gemma/gemma-skills) | 17 | Google agent skill patterns |
-#### RavenX Internal (Self-Extraction)
-| Source | Examples | Content |
-|--------|----------|---------|
-| RavenX-Sec repo (tools, docs) | 120 | RATH protocol, LEWM security, deployment guides |
----
-## Model Architecture
-```
-Layer 1: Qwen3.6-35B-A3B          ← 35B MoE brain (3B active per token)
-Layer 2: Claude 4.7 Opus distill   ← Enhanced reasoning pre-baked
-Layer 3: Abliteration              ← Zero refusals for security topics
-Layer 4: RavenX-Sec LoRA           ← 675,696 security examples (RATH)
-         ═══════════════════
-         RavenX-Sec v5.0           ← Vision + Security + Reasoning
-```
----
-## Training Details
-| Parameter | Value |
-|-----------|-------|
-| **Architecture** | Qwen3.6-35B-A3B (MoE) |
-| **Total Parameters** | 34.66B |
-| **Active Parameters** | ~3B per token |
-| **Trainable (LoRA)** | 64.1M (0.185%) |
-| **Training Data** | 675,696 examples |
-| **HuggingFace Datasets** | 21 sources (610,220 examples) |
-| **GitHub Repos** | 19 sources (65,476 examples) |
-| **LoRA Rank** | 32 |
-| **LoRA Layers** | 4 |
-| **Learning Rate** | 1e-5 |
-| **Batch Size** | 1 |
-| **Max Seq Length** | 1024 |
-| **Iterations** | 2,000 |
-| **Hardware** | Apple M4 Max 128GB |
-| **Peak Memory** | ~110GB |
----
-## The RavenX Model Family
-| Model | Domain | Protocol | Params | Training Data |
-|-------|--------|----------|--------|--------------|
-| [**RavenX-Sec v5.0**](https://huggingface.co/deadbydawn101/RavenX-Sec-35B-Security-RATH-mlx) | Security | 6-step RATH | 35B MoE | 675K (21 HF + 19 repos) |
-| [**RavenX-Sec v4.0**](https://huggingface.co/deadbydawn101/RavenX-Sec-8B-Security-RATH-128k-mlx-4bit) | Security | 6-step RATH | 8B | 610K |
-| [**RavenX-Trade v1.1**](https://huggingface.co/deadbydawn101/RavenX-Trade-8B-MAP-128k-mlx-4bit) | Trading | 4-step MAP | 8B | 318K |
----
-## Source Code & Training Pipeline
-**[github.com/DeadByDawn101/RavenX-Sec](https://github.com/DeadByDawn101/RavenX-Sec)**
-## License
-Apache-2.0
----
-*"We don't give up. We do what others don't and build what isn't possible." — RavenX LLC*

 ---
 license: apache-2.0
 language:
+- en
 library_name: mlx
+pipeline_tag: text-generation
+base_model: huihui-ai/Huihui-Qwen3.6-35B-A3B-Claude-4.7-Opus-abliterated
+tags:
+- text-generation
+- reasoning
+- distillation
+- chain-of-thought
+- qwen
+- qwen3.6
+- mixture-of-experts
+- moe
+- lora
+- unsloth
+- abliterated
+- uncensored
+- mlx
 ---