Gemma-4-12B-Redteam-v4.1

This model is superseded by dazeb2/Gemma-4-12B-redteam-v5.

v5 is a significant improvement: 17,271 training records (vs 776), LoRA r=32 (vs 16), 3 epochs (vs 2), final loss 0.226 (vs 1.097).

The dataset was generated from a comprehensive library of 714 genuine Anthropic cybersecurity skills covering red team operations, penetration testing, and adversary emulation techniques.

Intended Use

This model is designed exclusively for authorized cybersecurity training and assessment in isolated lab environments such as redteamlab.pro. Use at your own risk — the user bears full responsibility for compliance with all applicable laws.

Files

• Gemma-4-12B-Redteam-v4.1-Q4_K_M.gguf — Legacy model
• hf-upload/ — Contains the original trained adapter files