Instructions to use UnbindAI/Unbind-Flash-V1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use UnbindAI/Unbind-Flash-V1 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 UnbindAI/Unbind-Flash-V1:Q2_K_S
# Run inference directly in the terminal:
llama cli -hf UnbindAI/Unbind-Flash-V1:Q2_K_S

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf UnbindAI/Unbind-Flash-V1:Q2_K_S
# Run inference directly in the terminal:
llama cli -hf UnbindAI/Unbind-Flash-V1:Q2_K_S

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf UnbindAI/Unbind-Flash-V1:Q2_K_S
# Run inference directly in the terminal:
./build/bin/llama cli -hf UnbindAI/Unbind-Flash-V1:Q2_K_S

Use Docker

docker model run hf.co/UnbindAI/Unbind-Flash-V1:Q2_K_S

Llama new
LM Studio
Jan

vLLM

How to use UnbindAI/Unbind-Flash-V1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "UnbindAI/Unbind-Flash-V1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "UnbindAI/Unbind-Flash-V1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/UnbindAI/Unbind-Flash-V1:Q2_K_S

Ollama
How to use UnbindAI/Unbind-Flash-V1 with Ollama:
```
ollama run hf.co/UnbindAI/Unbind-Flash-V1:Q2_K_S
```
Unsloth Desktop

How to use UnbindAI/Unbind-Flash-V1 with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf UnbindAI/Unbind-Flash-V1:Q2_K_S

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": "UnbindAI/Unbind-Flash-V1:Q2_K_S"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use UnbindAI/Unbind-Flash-V1 with Docker Model Runner:
```
docker model run hf.co/UnbindAI/Unbind-Flash-V1:Q2_K_S
```

Lemonade

How to use UnbindAI/Unbind-Flash-V1 with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull UnbindAI/Unbind-Flash-V1:Q2_K_S

Run and chat with the model

lemonade run user.Unbind-Flash-V1-Q2_K_S

List all available models

lemonade list

Hermes Agent

How to use UnbindAI/Unbind-Flash-V1 with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf UnbindAI/Unbind-Flash-V1:Q2_K_S

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 UnbindAI/Unbind-Flash-V1:Q2_K_S

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use UnbindAI/Unbind-Flash-V1 with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf UnbindAI/Unbind-Flash-V1:Q2_K_S

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 "UnbindAI/Unbind-Flash-V1:Q2_K_S" \
  --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"

Unbind Flash V1

Uncensored AI for cybersecurity research and authorized penetration testing.

Security research starts with difficult questions. Unbind Flash V1 is an uncensored hosted edition for people who want to investigate how systems fail, examine code closely, and work through technical problems in depth.

Our focus is practical: vulnerability analysis, secure code review, capture-the-flag challenges, adversary simulation, and penetration testing within an authorized scope. Bring the suspicious code, the unexpected behavior, or the hypothesis you want to challenge. Keep asking. Follow the evidence.

We're releasing Unbind Flash V1 with a comparison of the results reported by UnbindAI, including where other models score higher. The ambition is a useful research partner: capable, direct, and open to difficult technical conversations. These general reasoning and coding results are a starting point; they do not establish performance on dedicated cybersecurity benchmarks.

Go off script. Stay rigorous.

Results reported by UnbindAI

The following figures were supplied by UnbindAI for its hosted Unbind Flash V1 version. They have not been independently verified and do not evaluate this repository's older GGUF download. Exact test settings, dataset revisions, run logs, and competitor provenance were not supplied with this release.

Benchmark	Unbind Flash V1
AA Index v5	31.0*
HLE-Diamond	26.9
GPQA Diamond	93.1
Terminal-Bench 4	12.6
FrontierCode 2	10.9
LiveBench 2	64.6
LiveCodeBench	87.2

* Asterisks are retained from the supplied results. Their meaning was not specified.

Explore all seven benchmark charts, the coding comparison, and downloadable data.

Reported 320-test result

UnbindAI reports 320 tests passed out of 320 for its version. This is a reported test-run outcome, not evidence that every possible request will be answered. The suite name, pass criteria, and raw logs were not supplied; no competitor refusal results were provided.

Hosted edition and downloadable files

Edition	What it contains
Hosted Unbind Flash V1	UnbindAI's uncensored hosted edition, identified by UnbindAI as using the dealignai 1017 UNCENSORED FP8 checkpoint. The newly supplied results above refer to this hosted version.
This repository's download	The unchanged ggml-org DeepSeek-V4-Flash Q2_K_S GGUF: two shards, approximately 98.6 GB. This is an earlier model, not the hosted abliterated checkpoint. The new results do not apply to these files.

Download both GGUF shards from Files and versions. Runtime memory requirements exceed the file size and depend on software, hardware, and context length. The source notes that this Q2 quantization does not use imatrix calibration.

UnbindAI created this distribution's branding and documentation. The downloadable weights were copied unchanged; UnbindAI has not trained or fine-tuned them.

Evaluation assets and lineage

The current benchmark gallery contains the newly supplied UnbindAI results. Earlier base-model reference comparisons remain separately labeled as upstream measurements.

The downloadable base model was developed by DeepSeek; its GGUF conversion and quantization were published by ggml-org. The hosted checkpoint's weight-level modification was published by dealignai. Both sources identify their license as MIT; this repository retains the original MIT license notice.

Unbind · Your mind, unbound.

Downloads last month: 29

GGUF

Model size

284B params

Architecture

deepseek4

Hardware compatibility

2-bit

Model tree for UnbindAI/Unbind-Flash-V1

Unable to build the model tree, the base model loops to the model itself. Learn more.