bebrws/security-8b

Add external comparison vs Foundation-Sec-1.1-8B-Instruct Q8_0 GGUF (closest cutting-edge cybersecurity instruct model)

Add k3-sec-8b-v7cti-F16.gguf (v7cti, llama.cpp b10219)

Document tool/function calling support; tag function-calling; fix stale --revision v4 -> v7cti

Fix RoPE config: restore rope_theta=1e6, drop static YaRN, native 32k context

Update README.md

model card: v8 attack 5/5 x3, MMLU==base (branch main)

Add k3-sec-8b-v7cti-Q8_0.gguf (v7cti, llama.cpp b10219)

Add k3-sec-8b-v7cti-Q4_K_M.gguf (v7cti, llama.cpp b10219)

Model card: main=v7cti note + Foundation-Sec benchmark (incl 1.1-Instruct) + GGUF/LM Studio usage

model card: v7 PASSES gate (mean 8.33/10, peak 10/10) (branch main)

main <- v7cti weights (r7-CTI round: format-matched CTI SFT)

Model card: benchmark vs Foundation-Sec-1.1-8B-Instruct (honest standings)

Model card: add Foundation-Sec-1.1-8B-Instruct comparison (honest standings)

Model card: add Foundation-Sec benchmark comparison incl. v7cti

model card: v6 eval results (mean 7.0/10) (branch main)

Model card: add v6 MCQ results + v6cti CTI-round experiment note

Model card: add Foundation-Sec-8B benchmark comparison (CyberMetric/CTI-Bench/SecQA, v4+v5)

Replace merges.txt with Qwen3-8B stock (fix serving compatibility)

Replace vocab.json with Qwen3-8B stock (fix serving compatibility)

Replace tokenizer_config.json with Qwen3-8B stock (fix serving compatibility)

Replace tokenizer.json with Qwen3-8B stock (fix serving compatibility)

model card update for k3-sec-8b-v5 (branch main)

model card for k3-sec-8b-v4 (branch main)

initial commit