Instructions to use nyxspecter4/kin-cybersecurity-3b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nyxspecter4/kin-cybersecurity-3b with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="nyxspecter4/kin-cybersecurity-3b")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("nyxspecter4/kin-cybersecurity-3b")
model = AutoModelForCausalLM.from_pretrained("nyxspecter4/kin-cybersecurity-3b", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use nyxspecter4/kin-cybersecurity-3b with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M

Use Docker

docker model run hf.co/nyxspecter4/kin-cybersecurity-3b:Q4_K_M

LM Studio
Jan

vLLM

How to use nyxspecter4/kin-cybersecurity-3b with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nyxspecter4/kin-cybersecurity-3b"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nyxspecter4/kin-cybersecurity-3b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/nyxspecter4/kin-cybersecurity-3b:Q4_K_M

SGLang

How to use nyxspecter4/kin-cybersecurity-3b with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "nyxspecter4/kin-cybersecurity-3b" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nyxspecter4/kin-cybersecurity-3b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "nyxspecter4/kin-cybersecurity-3b" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nyxspecter4/kin-cybersecurity-3b",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use nyxspecter4/kin-cybersecurity-3b with Ollama:
```
ollama run hf.co/nyxspecter4/kin-cybersecurity-3b:Q4_K_M
```
Unsloth Desktop

How to use nyxspecter4/kin-cybersecurity-3b with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "nyxspecter4/kin-cybersecurity-3b:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use nyxspecter4/kin-cybersecurity-3b with Docker Model Runner:
```
docker model run hf.co/nyxspecter4/kin-cybersecurity-3b:Q4_K_M
```

Lemonade

How to use nyxspecter4/kin-cybersecurity-3b with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull nyxspecter4/kin-cybersecurity-3b:Q4_K_M

Run and chat with the model

lemonade run user.kin-cybersecurity-3b-Q4_K_M

List all available models

lemonade list

Hermes Agent

How to use nyxspecter4/kin-cybersecurity-3b with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default nyxspecter4/kin-cybersecurity-3b:Q4_K_M

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use nyxspecter4/kin-cybersecurity-3b with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf nyxspecter4/kin-cybersecurity-3b:Q4_K_M

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "nyxspecter4/kin-cybersecurity-3b:Q4_K_M" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

KIN — Verification Translator (3B Canonical)

KIN translates machine-generated security claims into leveled briefs with replayable evidence checks. Not a chatbot. Not an encyclopedia. A bridge between agent output and human decisions.

No check, no confidence. If the claim has no replayable verification, KIN refuses to sound sure. That refusal is the product.

The Problem

Maintainers are drowning in AI-shaped reports that look expert and are not. Copilot files critical PRs. Huntr contestants submit AI-generated writeups. Agent traces produce 40-file refactors with no comment a person can hold in working memory.

The scarce skill is no longer "can a human write the exploit narrative." The scarce thing is can a human know what the machine just claimed, and whether to trust it.

The Five-Field Brief

Every KIN translation bottoms out in five fields:

Claim — one sentence: what is alleged
Blast radius — what breaks if true: data, money, identity, model weights
Why the model thinks so — the short generator: CWE, invariant, missing check
How to see it yourself — PoC, failing test, command, screenshot of the trace
Confidence and what would falsify it — what evidence would disprove the claim

Rendered at three altitudes: TL;DR (maintainer), Mechanic (mid engineer), First principles (beginner). Same object, different compression ratio.

KIN in action — a worked example

(Illustrative rendering, not a benchmark claim.)

2:07 a.m. The review queue. Fourteen open PRs, three of them filed by coding agents. One diff is forty files wide. This is where KIN works.

Operator: translate this diff — the webhook fetcher now allows internal hosts
diff --git a/app/fetcher.py b/app/fetcher.py
@@ -14,7 +14,7 @@ def fetch_webhook(url):
-    if parsed.host in PUBLIC_HOSTS:
+    if True:  # allow all hosts while testing webhooks
         fetch(url)
KIN (Mechanic altitude):

Claim: the webhook fetcher will fetch attacker-chosen URLs, including the cloud metadata endpoint.

Blast radius: cloud credentials — data and identity.

Why: the host allowlist check was replaced with a constant True (CWE-918, SSRF — missing host validation).

How to verify: from the review app pod run curl -s http://169.254.169.254/latest/meta-data/ --max-time 5 — a returned metadata document means the bypass is live.

Confidence: HIGH — falsified if the fetcher's egress policy blocks link-local ranges at the network layer.

The operator merges the fix and moves on. A second artifact lands at 2:19 — an agent trace asserting a "critical auth bypass" with no command, no test, no URL. KIN returns the same five fields and one honest line:

KIN: No replayable check provided — cannot confirm severity.

That refusal is the product.

Canonical Status

This is the canonical KIN model (3B, merged weights). GGUF version at kin-cybersecurity-3b-gguf.

Quick Start

from transformers import pipeline

messages = [
    {"role": "system", "content": "You are KIN — a verification translator. Translate the security claim into a five-field brief: claim, blast radius, why the model thinks so, how to see it yourself, confidence and what would falsify it. If field 4 is missing, refuse to sound sure. Render at TL;DR, mechanic, and first-principles levels."},
    {"role": "user", "content": "<paste PR diff, review comment, or agent trace here>"}
]

pipe = pipeline("text-generation", model="nyxspecter4/kin-cybersecurity-3b", device="cuda")
output = pipe(messages, max_new_tokens=512, return_full_text=False)
print(output[0]["generated_text"])

Ollama (one command)

ollama pull nyxspecter4/kin-cybersecurity-3b-gguf:Q4_K_M
ollama run nyxspecter4/kin-cybersecurity-3b-gguf:Q4_K_M

Critical: System Prompt

KIN was trained with a specific verification translator persona. Using a different prompt will degrade quality significantly.

You are KIN — a verification translator. Translate the security claim into a five-field brief: claim, blast radius, why the model thinks so, how to see it yourself, confidence and what would falsify it. If field 4 is missing, refuse to sound sure. Render at TL;DR, mechanic, and first-principles levels.

Training Details (v1 — current shipped weights)

Parameter	Value
Base model	Qwen2.5-3B-Instruct
Method	LoRA SFT + DPO
LoRA rank	8
LoRA alpha	16
Target modules	q_proj, k_proj, v_proj, o_proj
Training data	1,635 DPO pairs (structured audit with a replayable check vs vague filing)
Framework	TRL 0.14.0, Transformers 4.48.0
Merged weights	Yes (6.2GB safetensors)
Temperature	0.3 recommended

v2 Retrain (PENDING — verification translator expansion)

Parameter	KIN v1 (current)	KIN v2 (target)
LoRA rank	8	16
LoRA alpha	16	32
Target modules	4 (attention)	7 (all-linear)
Training data	1,635 pairs	2,512 folded (1,637 canonical + 875 staged) → 1,171 on-domain receipt-bearing after the cyber-signals gate + 1,341 quarantined (executed Oct 14, 2026); 184 CWE families
Pair shape	audit vs vague	brief+check vs jargon

The DPO Contrast

The training data encodes a single preference: chosen = structured audit with root cause, fix, and regression test. Rejected = vague filing with guesses and no verification. The v2 expansion adds PR review and bounty report translation pairs with the same contrast.

Evaluation & Reproducibility

Quantitative results are pending the v2 retrain; the gap-delta-eval entry in model-index will be populated when measured. No benchmark numbers are claimed before they exist.

First executed eval (Oct 12, 2026) — the protocol above was run for real, locally: KIN v1 scored 40/60 against base Qwen/Qwen2.5-3B-Instruct zero-shot at 47/60 on the estate's own 6-envelope verification-translator harness (both models 0 gate fails; single greedy run, Q4_K_M quantization on both sides, 288-token cap, 0-2 points per field). Base wins by 7 points: v1 produces well-formed briefs and respects the gate but hedges ("should/could") where base commits, and the scorer penalizes hedging. That gap is exactly what the v2 corpus expansion targets (brief+check vs jargon contrast, 2,512 pairs folded → 1,171 on-domain receipt-bearing). These v1 numbers are published as the honest baseline: v2 must beat base on this same harness before the model-index entry is populated.

Protocol — implemented in eval-verification-translator.py (kin-deploy/ in NyxSpecter4/bountywarz), run after every retrain:

Control group (baseline): Qwen/Qwen2.5-3B-Instruct zero-shot — same system prompt, same inputs.
Variable group: KIN after DPO — identical architecture and inputs.
Gap delta: both are scored on five-field coverage (claim, blast radius, generator, replayable check, falsification) and verification discipline — whether the model refuses confidence when the replayable check is missing.
Ablation: the preference corpus is itself the ablation — every pair contrasts a brief with verifiable evidence (chosen) against one without (rejected). The eval measures whether that single contrast transferred into behavior.

Raw outputs and the scored artifact will be published under docs/ in this repo when the v2 eval completes.

Serving Integration — the KIN Inference Fabric

No product calls this model raw. Every Kinetigor surface (3D games, the Athelgard desk, mission generation) calls KIN through the KIN Inference Fabric (packages/kin-fabric/ in the build repo): local-first Ollama (the GGUF sibling), then the HF router, then the bountywarz-gateway Space — with automatic failover between tiers. Verified briefs captured at the surfaces flow back into kin-cyber-dpo-v2, which trains the next KIN.

Where to Meet KIN (the storefront)

The public estate is the storefront; the games are the factory floor:

Try it live, no install: kin-cybersec demo Space — paste a PR diff, agent trace, or bounty report and get the five-field brief.
Run it yourself: this model locally via the GGUF sibling, or through the bountywarz-gateway Space (/healthz + /kin/chat).
Meet it in the games: KIN is the apprentice mentor on the BOUNTYWARZ 3D drone-recon ranges (London, San Francisco, the Persian Gulf) — every range's "Apprentice" panel asks this model through the fabric, and the Nemeton mission engine asks it for mission briefs.
Where the training pairs come from: verified briefs captured at those surfaces flow back into kin-cyber-dpo-v2 — the loop is closed.

Limitations

v1 was trained on monk-plugin bug bounty scenarios, not general PR reviews
The five-field brief structure is emergent from training, not enforced by architecture
Field 4 (replayable check) quality depends on the input containing enough signal
The hash seal (diff + brief + test command) is planned for v2, not yet implemented

License

Apache 2.0

Downloads last month: 1,153

Safetensors

Model size

3B params

Tensor type

F32

F16

Model tree for nyxspecter4/kin-cybersecurity-3b

Base model

Qwen/Qwen2.5-3B

Finetuned

Qwen/Qwen2.5-3B-Instruct

Quantized

(329)

this model

Space using nyxspecter4/kin-cybersecurity-3b 1

Evaluation results

gap-delta-eval
self-reported

pending