← Greps
- HF Model
- hotdogs/qwen3.6-27b-cybersecurity-lora
- Base Model
- Qwen/Qwen3.6-27B
- Pipeline
- text-generation
- Downloads
- 517
- Likes
- 9
- cybersecurity
- lora
- qwen3.6
- sft
- fenrir
- peft
cyber-llm
Qwen3.6-27B Cybersecurity LoRA (Fenrir SFT)
Uploaded to Hugging Face by hotdogs — PEFT LoRA on Qwen3.6-27B trained with AlicanKiraz0 Fenrir cybersecurity dataset; GGUF tags present.
What Is This?
Community upload on Hugging Face by hotdogs. qwen3.6-27b-cybersecurity-lora is a PEFT LoRA adapter on Qwen/Qwen3.6-27B, supervised-fine-tuned with the AlicanKiraz0/Cybersecurity-Dataset-Fenrir-v2.1 corpus (same Fenrir lineage as other cyber LLMs already on Greps). Tags include gguf for local inference packaging. It is not an OFFSITE.DARK release.
Created mid-June 2026; last modified 2026-07-05 in this indexing window.
Metadata
| Field | Value |
|---|---|
| Author | hotdogs |
| Base model | Qwen/Qwen3.6-27B |
| Method | LoRA / SFT (PEFT) |
| Dataset | AlicanKiraz0/Cybersecurity-Dataset-Fenrir-v2.1 |
| Pipeline | text-generation |
| License | Apache-2.0 |
| Downloads | ~517 |
| Likes | 9 |
Why Index It
- Continues the Fenrir/Qwen cybersecurity fine-tune cluster already represented by Titus / BaronLLM-adjacent greps.
- 27B class is large enough for serious local cyber-assistant experiments.
- Useful contrast to smaller Strix XSS / RavenX agent greps.
Caveats
LoRA adapters inherit base-model jailbreak and hallucination risks. Use only in authorized research sandboxes; do not treat outputs as authoritative vuln intel.