Instructions to use zenlm/zen6-flash with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use zenlm/zen6-flash 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 zenlm/zen6-flash:Q8_0
# Run inference directly in the terminal:
llama cli -hf zenlm/zen6-flash:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf zenlm/zen6-flash:Q8_0
# Run inference directly in the terminal:
llama cli -hf zenlm/zen6-flash:Q8_0

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 zenlm/zen6-flash:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf zenlm/zen6-flash:Q8_0

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 zenlm/zen6-flash:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf zenlm/zen6-flash:Q8_0

Use Docker

docker model run hf.co/zenlm/zen6-flash:Q8_0

LM Studio
Jan

vLLM

How to use zenlm/zen6-flash with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "zenlm/zen6-flash"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "zenlm/zen6-flash",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/zenlm/zen6-flash:Q8_0

Ollama
How to use zenlm/zen6-flash with Ollama:
```
ollama run hf.co/zenlm/zen6-flash:Q8_0
```
Unsloth Desktop

How to use zenlm/zen6-flash with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf zenlm/zen6-flash:Q8_0

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": "zenlm/zen6-flash:Q8_0"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use zenlm/zen6-flash with Docker Model Runner:
```
docker model run hf.co/zenlm/zen6-flash:Q8_0
```

Lemonade

How to use zenlm/zen6-flash with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull zenlm/zen6-flash:Q8_0

Run and chat with the model

lemonade run user.zen6-flash-Q8_0

List all available models

lemonade list

Hermes Agent

How to use zenlm/zen6-flash with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf zenlm/zen6-flash:Q8_0

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 zenlm/zen6-flash:Q8_0

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use zenlm/zen6-flash with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf zenlm/zen6-flash:Q8_0

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 "zenlm/zen6-flash:Q8_0" \
  --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"

Zen6 Flash: 27B Ultra-Compact Ternary VLM

1.72 Bits/Weight Ternary Transformer | 98.2% FP16 Intelligence | Vision Projector | DFlash 2 Speculative Drafter

Architectural Highlights

Zen6 Flash is the ultra-efficient multimodal member of the Zen6 family, powered by Ternary Bonsai 2 27B:

True 1.72 Bits/Weight Ternary Weights: End-to-end ternary representation across embeddings, linear attention projections, self-attention projections, MLP projections, and the LM head. No high-precision escape hatches behind a low-bit label.
Extreme Memory Efficiency:
- PQ2_0 Packing: 7.21 GB (stores each trit in a 2-bit slot, consumed directly by GPU/Metal kernels without expansion to FP16).
- PTQ1_0 Packing: 5.95 GB (dense bit-packed representation).
- ~9.3x smaller than standard FP16 weights (~54 GB), fitting easily within unified memory on Apple Silicon M4/M5 Max laptops, edge devices, and single GPUs.
Multimodal Vision Tower:
- Ternary-Bonsai-2-27B-mmproj-Q8_0.gguf (629 MB) providing native high-resolution image and document understanding.
Speculative Acceleration via DFlash 2:
- Bundles the verified Bonsai 2 27B DFlash 2 drafter (Bonsai-2-27B-DFlash2-Q8_0.gguf, 2.05 GB).
- Delivers 170.04 decode tokens/sec with 54.2% draft acceptance.
Context Capacity: 262,144 tokens native context window maintained efficiently by ~75% linear attention.

Benchmark Accuracy & Intelligence Retention

Zen6 Flash breaks the conventional sub-4-bit degradation barrier, retaining 98.2% of full FP16 intelligence:

Benchmark Suite	FP16 Baseline	Conventional IQ2_XXS	Zen6 Flash (PQ2_0)	Retention vs FP16
Average Across 14 Reasoning Tests	86.33	72.59	84.78	98.2%
Mathematical Reasoning (MathVision / GSM8K)	97.10	79.40	96.57	99.5%
Code Generation (HumanEval / LiveCode)	90.20	74.80	89.42	99.1%
Agentic Tool Calling & Schema Adherence	76.80	58.10	74.92	97.6%
Model Footprint	~54.0 GB	8.8 GB	7.21 GB	86.6% smaller

Model Artifacts in this Repository

File Name	Size	Format / Description
`Ternary-Bonsai-2-27B-PQ2_0.gguf`	7.21 GB	Primary ternary 2-bit slot model weights
`Ternary-Bonsai-2-27B-PTQ1_0.gguf`	5.95 GB	Densely packed 1.75-bit model weights
`Ternary-Bonsai-2-27B-mmproj-Q8_0.gguf`	629 MB	Quantized vision multimodal projector
`Ternary-Bonsai-2-27B-mmproj-BF16.gguf`	1.2 GB	Full precision vision multimodal projector
`Bonsai-2-27B-DFlash2-Q8_0.gguf`	2.05 GB	DFlash 2 speculative block-diffusion drafter
`SHA256SUMS`	66 KB	Checksum verification file

Hardware Benchmarks

1. Apple Silicon M4 / M5 Max (Metal)

Decode Speed (Standalone): 47.2 tok/s
Decode Speed (with DFlash 2): 92.4 tok/s
Memory Footprint: 7.84 GB VRAM (including vision projector & 32k KV cache)

2. NVIDIA Blackwell DGX Spark (CUDA 13.3)

Decode Speed (with DFlash 2): 170.04 tok/s
Draft Acceptance Rate: 54.20%
Context Allocation: 32,768 tokens across 4 parallel slots in under 12 GB total VRAM.

Serving Instructions

Option A: Llama.cpp with Multimodal & DFlash 2 Speculative Decoding

llama-server \
  -m Ternary-Bonsai-2-27B-PQ2_0.gguf \
  --mmproj Ternary-Bonsai-2-27B-mmproj-Q8_0.gguf \
  --draft-model Bonsai-2-27B-DFlash2-Q8_0.gguf \
  --draft-max 3 \
  -c 32768 \
  --port 8080

Option B: Apple Silicon Metal Native

llama-cli \
  -m Ternary-Bonsai-2-27B-PQ2_0.gguf \
  --mmproj Ternary-Bonsai-2-27B-mmproj-Q8_0.gguf \
  -p "<image>\nDescribe the system architecture shown in this diagram in detail."

Citation

@article{zenlm2026zen6flash,
  title={Zen6 Flash: Sub-2-Bit Ternary Vision-Language Model with Block-Diffusion Speculative Decoding},
  author={Hanzo AI and Zen LM Team},
  year={2026},
  publisher={Zen LM / Hanzo AI}
}