Qwen3-4B-Instruct-2507 fine-tuned on an offensive-security SFT dataset (1,226 rows). Elite-hacker persona on casual queries, structured markdown methodology on technical ones.

Files

qwen3-4b-instruct-2507.Q4_K_M.gguf - quantized weights (~2.5 GB)
Modelfile - Ollama template with correct ChatML stop tokens + Zero Stack system prompt

Run with Ollama

bash

ollama create zerostack-4b -f Modelfile
ollama run zerostack-4b

Run with llama.cpp

./llama-cli -m qwen3-4b-instruct-2507.Q4_K_M.gguf -p "hello"

Training

Base: Qwen3-4B-Instruct-2507
Method: LoRA (r=32), 3 epochs, Unsloth
Dataset: SFT_GENERALIST (1,226 rows, ChatML)

Intended Use

Authorized security testing, CTF practice, red-team research, and security education. Targeted at practitioners who already know what they're doing and want fast recall of commands, workflows, and methodology.

Limitations & Risks

May hallucinate specific CVE IDs, tool flags, or payload syntax - verify against primary sources before running.
No safety guardrails against misuse. Do not use against systems you don't own or have explicit written authorization to test.
Small model (4B) - shallower reasoning than the 14B; prefer 14B for multi-step planning.
Persona responses are stylistic flavor, not a safety signal.
Trained on English data only; non-English performance is not evaluated.

License / Use

For authorized security testing, research, and educational use only. Do not use for unauthorized access to systems you do not own or have explicit permission to test.

redstackio/qwen3-4b-redstack-v1

Zero Stack - Qwen3-4B (GGUF, Q4_K_M)