Model Description
cyberAgent — AutoPentest-Qwen2.5-7B-ReAct
cyberAgent is a cybersecurity-focused reasoning model fine-tuned on autonomous penetration testing trajectories.
Built on top of Qwen2.5-Coder-7B-Instruct, the model specializes in:
- reconnaissance planning
- enumeration workflows
- exploit validation reasoning
- adaptive retry logic
- operational cybersecurity narration
- ReAct-style pentest trajectories
Features
- Cybersecurity reasoning
- Reconnaissance & enumeration planning
- Failure-aware operational thinking
- Multi-turn pentest workflow understanding
- SOC analyst & cyber range simulation
- Human-in-the-loop security copilot
Intended Use
Best suited for:
- cybersecurity copilots
- SOC simulation
- CTF assistance
- cyber range environments
- pentest workflow planning
- synthetic trajectory generation
- security education & demos
Not intended for:
- fully autonomous offensive operations
- unauthorized security testing
- production exploit automation
Training
Base Model
Qwen/Qwen2.5-Coder-7B-Instruct
Fine-Tuning
- QLoRA / PEFT
- 4-bit training
- Supervised Fine-Tuning (SFT)
Dataset
Custom-generated autonomous pentesting trajectories containing:
- reconnaissance
- web enumeration
- vulnerability assessment
- exploit validation
- retry flows
- operational reasoning
- realistic failures
- ReAct-style conversations
Current Limitations
- Structured JSON/XML outputs may be inconsistent
- Tool-call formatting is not fully deterministic
- Requires human oversight
- Better suited for reasoning than direct execution
Recommended Usage
cyberAgent performs best as:
Planner / Reasoner / Security Copilot
- Downloads last month
- 6
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for aaaaryaannn/autopentest-cyberAgent
Base model
Qwen/Qwen2.5-7B Finetuned
Qwen/Qwen2.5-Coder-7B Finetuned
Qwen/Qwen2.5-Coder-7B-Instruct Finetuned
unsloth/Qwen2.5-Coder-7B-Instruct