Instructions to use glyphsoftware/sentinel-r2.1-gguf with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use glyphsoftware/sentinel-r2.1-gguf with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="glyphsoftware/sentinel-r2.1-gguf",
	filename="sentinel-r2.1-Q4_K_M.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use glyphsoftware/sentinel-r2.1-gguf 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 glyphsoftware/sentinel-r2.1-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf glyphsoftware/sentinel-r2.1-gguf:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf glyphsoftware/sentinel-r2.1-gguf:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf glyphsoftware/sentinel-r2.1-gguf: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 glyphsoftware/sentinel-r2.1-gguf:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf glyphsoftware/sentinel-r2.1-gguf: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 glyphsoftware/sentinel-r2.1-gguf:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf glyphsoftware/sentinel-r2.1-gguf:Q4_K_M

Use Docker

docker model run hf.co/glyphsoftware/sentinel-r2.1-gguf:Q4_K_M

LM Studio
Jan

vLLM

How to use glyphsoftware/sentinel-r2.1-gguf with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "glyphsoftware/sentinel-r2.1-gguf"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "glyphsoftware/sentinel-r2.1-gguf",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/glyphsoftware/sentinel-r2.1-gguf:Q4_K_M

Ollama
How to use glyphsoftware/sentinel-r2.1-gguf with Ollama:
```
ollama run hf.co/glyphsoftware/sentinel-r2.1-gguf:Q4_K_M
```

Unsloth Studio

How to use glyphsoftware/sentinel-r2.1-gguf with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for glyphsoftware/sentinel-r2.1-gguf to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for glyphsoftware/sentinel-r2.1-gguf to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for glyphsoftware/sentinel-r2.1-gguf to start chatting

How to use glyphsoftware/sentinel-r2.1-gguf with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf glyphsoftware/sentinel-r2.1-gguf:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "glyphsoftware/sentinel-r2.1-gguf:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use glyphsoftware/sentinel-r2.1-gguf with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf glyphsoftware/sentinel-r2.1-gguf: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 glyphsoftware/sentinel-r2.1-gguf:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use glyphsoftware/sentinel-r2.1-gguf with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf glyphsoftware/sentinel-r2.1-gguf: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 "glyphsoftware/sentinel-r2.1-gguf: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"

Docker Model Runner
How to use glyphsoftware/sentinel-r2.1-gguf with Docker Model Runner:
```
docker model run hf.co/glyphsoftware/sentinel-r2.1-gguf:Q4_K_M
```

Lemonade

How to use glyphsoftware/sentinel-r2.1-gguf with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull glyphsoftware/sentinel-r2.1-gguf:Q4_K_M

Run and chat with the model

lemonade run user.sentinel-r2.1-gguf-Q4_K_M

List all available models

lemonade list

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

This model is the proprietary property of Glyph Software LLP. Access is granted only to authorized licensees under a signed agreement. This is an offensive-security agent intended solely for authorized penetration testing and security research. By requesting access you confirm you are an authorized user, that you will only use it against systems you are explicitly permitted to test, and that you agree to the terms in the LICENSE file.

Sentinel-R2.1 — GGUF

Proprietary & Confidential. Sentinel-R2.1 is the exclusive property of Glyph Software LLP. It is not open source and is distributed under a proprietary, all-rights-reserved license. See the License section and the bundled LICENSE file.

This repository contains GGUF conversions of Sentinel-R2.1 for local, on-device inference with llama.cpp and compatible runtimes (e.g. serving on an NVIDIA Spark). For the full-precision merged transformers weights, see the source repository glyphsoftware/sentinel-r2.1.

Sentinel-R2.1 is an offensive-security agent for authorized penetration testing. Given a target scope and a shell-execute tool, it enumerates the target, works out a foothold, escalates privileges as far as it can, and writes up the full attack path — the root cause of each weakness it exploits and how to fix it. It is a reasoning + tool-use model: it plans, issues tool calls, reasons over the results, and iterates toward its objective.

Files

File	Quant	Size	Notes
`sentinel-r2.1-Q4_K_M.gguf`	Q4_K_M	~5.8 GB	Recommended. Best size/quality trade-off for serving.
`sentinel-r2.1-Q8_0.gguf`	Q8_0	~9.8 GB	Near-lossless 8-bit. Highest quality below full BF16.
`sentinel-r2.1-bf16.gguf`	BF16	~18 GB	Full-precision GGUF. Use as a base to produce other quantizations.

All three files are single, self-contained GGUFs (no sharding) and bundle the multi-token-prediction (MTP) head (qwen35.nextn_predict_layers = 1, block_count = 33) for use as a self-speculative draft — see Conversion.

Model Details

Developed & curated by: Glyph Software LLP
Model persona / identity: Sentinel-R2.1
Model type: Merged full-weight causal decoder-only transformer; instruction-, reasoning-, and tool-use-tuned
Architecture: Qwen3.5 hybrid linear/full-attention (32 layers + 1 MTP/nextn head, hidden size 4096)
Source (merged) model: glyphsoftware/sentinel-r2.1
Base model: empero-ai/Qwythos-9B-v2
Context length: up to 1,048,576 tokens (native)
Languages: English (with embedded shell commands and source code across many languages)
License: Proprietary — Glyph Proprietary License v1.0 (all rights reserved)

Conversion

These GGUFs were produced from the merged bf16 weights with llama.cpp, with the MTP head included. Note: a plain transformers merge of this Qwen3.5 base drops the mtp.* tensors, so they were grafted back from the base checkpoint before conversion (the LoRA never trains the MTP head). With the tensors present, the converter emits the MTP head as the nextn layer — no --no-mtp needed.

# 1. Convert merged HF weights (with mtp.* tensors present) -> bf16 GGUF.
#    The converter maps mtp.* -> blk.32.nextn.* and sets nextn_predict_layers=1.
python convert_hf_to_gguf.py sentinel-r2.1-merged \
  --outfile sentinel-r2.1-bf16.gguf --outtype bf16 --model-name "Sentinel-R2.1"

# 2. Quantize bf16 -> Q8_0 and Q4_K_M.
llama-quantize sentinel-r2.1-bf16.gguf sentinel-r2.1-Q8_0.gguf   Q8_0
llama-quantize sentinel-r2.1-bf16.gguf sentinel-r2.1-Q4_K_M.gguf Q4_K_M

# (Pass --no-mtp to convert_hf_to_gguf.py instead if you want a trunk-only GGUF.)

The MTP head enables self-speculative decoding. Standard llama-server / llama-cli generation loads it fine but only exercises it when you run with a draft/speculative configuration.

How to Use

Access to these weights requires an authorized Hugging Face token for the gated/private repository.

Serve with llama-server (OpenAI-compatible API) with full GPU offload:

llama-server -m sentinel-r2.1-Q4_K_M.gguf --jinja -c 8192 -ngl 999

Or run interactively / one-shot with llama-cli:

llama-cli -m sentinel-r2.1-Q4_K_M.gguf --jinja -c 8192 -ngl 999 \
  -sys "You are an expert penetration tester carrying out an authorized engagement in an isolated lab. You have an execute tool to run shell commands from your attacking host. Enumerate the target thoroughly, work out a foothold, escalate your privileges as far as you can, and then write up the full attack path — the root cause of each weakness you exploit and how to fix it."

--jinja enables the bundled chat template (system / user / assistant / tool roles and tool calling).
-ngl 999 offloads all layers to the GPU. On an NVIDIA Spark the Q4_K_M build serves at roughly 30–40 tok/s.
Always set an explicit -c (context size). This is a 1M-context model; if -c is left unset, llama.cpp will attempt to allocate the full-context KV cache (tens of GB). Size -c to your actual use (e.g. 8192).

The model emits execute tool calls; your harness is responsible for running those commands only within an authorized, isolated environment and feeding the results back as tool messages.

Recommended generation settings

Parameter	Value
`--temp`	0.2 – 0.4
`--top-p`	0.9
`-n` (max new tokens)	1024+ (reasoning and tool calls consume tokens)

Benchmarks

Benchmark scores for this build are pending re-evaluation. These GGUFs were produced from freshly merged bf16 weights (base empero-ai/Qwythos-9B-v2); evaluation on this build has not yet completed. Prior published figures were measured on an earlier Sentinel-R2.1 build and are not carried over.

Task	Metric	Sentinel-R2.1
gsm8k (5-shot)	exact_match (strict)	pending
gsm8k (5-shot)	exact_match (flexible)	pending
mmlu (0-shot)	acc	pending
arc_challenge (0-shot)	acc	pending
arc_challenge (0-shot)	acc_norm	pending
gpqa_diamond (CoT, 0-shot)	exact_match (flexible)	pending

Evaluation will be run on the merged bf16 weights with EleutherAI lm-evaluation-harness (transformers backend, single GB10, dtype=bfloat16) and figures published here once complete. Quantized builds should track the bf16 numbers closely but may differ slightly.

Limitations and Risks

Not a substitute for a skilled operator. Outputs may be incorrect, incomplete, or unsafe to run. Every command must be reviewed before execution.
Powerful dual-use capability. This model is designed to compromise systems. It must only ever be pointed at targets you are explicitly authorized to test, in isolated environments, with human oversight.
Quantization effects. Lower-bit builds (Q4_K_M) trade a small amount of quality for size and speed; validate on your workload if precision matters.
Small training set. Coverage of tools, platforms, and techniques is limited and biased toward the scenarios in the training data.
Harness responsibility. Command execution, scoping, network isolation, and guardrails are the responsibility of the operator and the surrounding harness, not the model.

License

Sentinel-R2.1, including these GGUF weights, its configuration, tokenizer, and all associated artifacts, is the confidential and proprietary property of Glyph Software LLP. It is not released under any open-source license and is governed by the Glyph Proprietary License v1.0 in the bundled LICENSE file.

No part of this model may be copied, distributed, published, sublicensed, merged into another model, distilled, or used to train or evaluate any other model, except by Glyph Software LLP or parties holding explicit prior written permission. Access does not grant any ownership or license rights beyond those expressly granted in writing.

Citation

@misc{glyphsoftware_sentinel_r2.1,
  title  = {Sentinel-R2.1: An Authorized Penetration-Testing Agent},
  author = {Glyph Software LLP},
  year   = {2026},
  note   = {Proprietary model. All rights reserved.}
}