Orionfold/SecurityLLM-GGUF

SecurityLLM GGUF

GGUF quantizations of ZySec-AI/SecurityLLM, verified end-to-end on the NVIDIA DGX Spark (GB10, 128 GB unified memory).

Spark-tested

Every Orionfold quant ships with a measurement quad on the NVIDIA DGX Spark (GB10, 128 GB unified memory): perplexity, sustained tok/s, thermal envelope, and CyberMetric (n=50, mcq_letter) accuracy. The numbers below are the actual run, not a wishlist.

Variant	Size	Perplexity (wikitext-2)	tok/s on Spark	CyberMetric (n=50, mcq_letter)
Q4_K_M	4.1 GB	7.400	47.7	40.0%
Q5_K_M	4.8 GB	7.314	40.0	38.0%
Q6_K	5.5 GB	7.313	35.0	36.0%
Q8_0	7.2 GB	7.307	30.3	36.0%
F16	13.5 GB	7.301	17.4	34.0%

Thermal envelope: sustained-load minutes before thermal throttle on a single GB10 = 5 min. Beyond this, expect tok/s degradation; the duty-cycle disclosure is per Orionfold's quant-card standard.

Variants

Variant	Recommended use
Q4_K_M	Best balance — fits comfortably in Spark unified memory at 70B; default pick.
Q5_K_M	Higher quality than Q4_K_M with modest size bump.
Q6_K	Near-lossless; recommended if memory headroom allows.
Q8_0	Effectively lossless; reach for this when quality matters more than throughput.
F16	Reference — no quantization. Use only for measurement / baseline.

How to run

Pull a variant:

huggingface-cli download Orionfold/SecurityLLM-GGUF model-Q5_K_M.gguf \
  --local-dir ./models/securityllm

Serve it via llama-server (OpenAI-compatible API):

llama-server -m ./models/securityllm/model-Q5_K_M.gguf \
  -c 4096 -ngl 99 -t 8 \
  --host 0.0.0.0 --port 8080

Or run in-process via llama-cpp-python:

from llama_cpp import Llama
llm = Llama(
    model_path="./models/securityllm/model-Q5_K_M.gguf",
    n_ctx=4096, n_gpu_layers=99, chat_format="zephyr",
)
out = llm.create_chat_completion(
    messages=[
        {"role": "user",
         "content": "What is the primary purpose of a key-derivation function (KDF)?\n\n"
                    "A) Generate public keys\n"
                    "B) Authenticate digital signatures\n"
                    "C) Encrypt data using a password\n"
                    "D) Transform a secret into keys and Initialization Vectors\n\n"
                    "Reply with only the single letter A, B, C, or D."}
    ],
    temperature=0.0,
)
print(out["choices"][0]["message"]["content"])

LM Studio and Ollama (via a Modelfile) load the GGUF directly with no additional setup.

Methods

Full methodology and Spark-side measurement protocol: Vertical-curator quants on Spark — SecurityLLM-GGUF + CyberMetric mini-eval.

Other Orionfold vertical curators

Same Spark-tested recipe across the curator-on-Spark series:

• finance-chat-GGUF — AdaptLLM finance-chat (Llama-2-7B lineage) for FinanceBench-shaped queries
• Saul-7B-Instruct-v1-GGUF — Equall Saul-7B legal-instruct for LegalBench-shaped queries
• II-Medical-8B-GGUF — Qwen3-8B + DAPO reasoning for MedMCQA-shaped queries

Each card lists its own measurement quad; the headline numbers are recorded as the actual sweep ran, never pre-corrected.

---

Published by Orionfold LLC · orionfold.com · Methods documented at ainative.business/field-notes.

Want to know when the next Orionfold vertical curator drops? Join the launch list at orionfold.com.