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

File size: 3,343 Bytes

7deda38

# Unbind Flash V1 — benchmark comparisons

Unbind Flash V1 launch graphics. The highlighted “Unbind Flash V1 · base model” row represents the unmodified DeepSeek-V4.1-Flash-1017 reference checkpoint, using scores reported by DeepSeek. These are base-model comparisons, not evaluations of Unbind’s endpoint, abliterated hosted checkpoint or older public GGUF.

Each chart compares models within one benchmark on a 0–100 axis. Evaluation setups differ across benchmarks; do not average these scores into an overall rating.

[Source table and evaluation notes](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017).

## Animated overview

![Eight benchmark comparisons](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/benchmark-overview.gif)

## DeepSWE v1.1

![DeepSWE v1.1 comparison](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/01-deepswe.png)

Mini-SWE-V2 harness. Publisher-reported benchmark scores.

## Terminal-Bench 4.0

![Terminal-Bench 4.0 comparison](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/02-terminal.png)

Claude Code harness. Publisher-reported benchmark scores.

## HLE without tools

![HLE without tools comparison](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/03-hle.png)

Full HLE figures only. GLM-5.3 omitted: the source gives a text-only subset score.

## ProgramBench

![ProgramBench comparison](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/04-program.png)

Publisher-reported benchmark scores. Additional per-row settings are not specified.

## NL2Repo-Bench

![NL2Repo-Bench comparison](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/05-nl2repo.png)

Publisher-reported benchmark scores. Additional per-row settings are not specified.

## Automation-Bench v1.0.6

![Automation-Bench v1.0.6 comparison](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/06-automation.png)

Publisher-reported benchmark scores. Additional per-row settings are not specified.

## Agents’ Last Exam

![Agents’ Last Exam comparison](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/07-agents.png)

Publisher-reported benchmark scores. Additional per-row settings are not specified.

## Apex Bench v1.1

![Apex Bench v1.1 comparison](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/08-apex.png)

Publisher-reported benchmark scores. Additional per-row settings are not specified.

## Data and reuse

[CSV](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/source-data.csv) · [JSON](https://huggingface.co/UnbindAI/Unbind-Flash-V1/resolve/main/assets/benchmarks/source-data.json). The accompanying SVG files provide editable vector exports.

All plotted values reproduce the source table. The highlighted display label is branded as “Unbind Flash V1 · base model”; its exact source model is retained in the footer and data files. Other model labels reproduce the source table. GLM-5.3 is omitted from HLE because only its text-only subset score is supplied. These are reported benchmark comparisons, not newly run evaluations.