Instructions to use gkraker04/Nanbeige4.2-3B-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Notebooks
Google Colab
Kaggle
Local Apps Settings

How to use gkraker04/Nanbeige4.2-3B-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 gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf gkraker04/Nanbeige4.2-3B-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 gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama cli -hf gkraker04/Nanbeige4.2-3B-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 gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf gkraker04/Nanbeige4.2-3B-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 gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M

Use Docker

docker model run hf.co/gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use gkraker04/Nanbeige4.2-3B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "gkraker04/Nanbeige4.2-3B-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": "gkraker04/Nanbeige4.2-3B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M

Ollama
How to use gkraker04/Nanbeige4.2-3B-GGUF with Ollama:
```
ollama run hf.co/gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M
```

Unsloth Studio

How to use gkraker04/Nanbeige4.2-3B-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 gkraker04/Nanbeige4.2-3B-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 gkraker04/Nanbeige4.2-3B-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for gkraker04/Nanbeige4.2-3B-GGUF to start chatting

How to use gkraker04/Nanbeige4.2-3B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf gkraker04/Nanbeige4.2-3B-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": "gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use gkraker04/Nanbeige4.2-3B-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 gkraker04/Nanbeige4.2-3B-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 gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use gkraker04/Nanbeige4.2-3B-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf gkraker04/Nanbeige4.2-3B-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 "gkraker04/Nanbeige4.2-3B-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 gkraker04/Nanbeige4.2-3B-GGUF with Docker Model Runner:
```
docker model run hf.co/gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M
```

Lemonade

How to use gkraker04/Nanbeige4.2-3B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull gkraker04/Nanbeige4.2-3B-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Nanbeige4.2-3B-GGUF-Q4_K_M

List all available models

lemonade list

Nanbeige4.2-3B GGUF

GGUF quantizations of Nanbeige4.2-3B, a compact 3B agentic model built on a Looped Transformer architecture.

These quants are built for buun-llama-cpp, the fork that added Nanbeige architecture support. They'll work with any GGUF-compatible runtime that supports Nanbeige models.

Produced with importance-matrix-guided quantization (llama-quantize --imatrix) from the bf16 source, ordered highest to lowest bits-per-weight.

Which Quant Should I Use?

Use Case	Recommended Quant	Size	Why
Maximum quality	Q8_0 or Q6_K	3.2–4.1 GB	Virtually lossless
Best tradeoff	Q4_K_M	2.4 GB	Sweet spot: near-lossless quality, tiny footprint
Memory constrained	Q3_K_M or IQ3_M	1.9–2.0 GB	Still very capable
Edge/mobile	IQ2_M or Q2_K	1.5–1.6 GB	Noticeable but functional
Maximum compression	Q1_0	0.87 GB	Fits anywhere, quality degrades significantly

Quantization Ladder

Quant	Size	bpw
bf16	7.77 GB	16.00
Q8_0	4.13 GB	8.00
Q6_K	3.19 GB	6.00
Q5_1	2.95 GB	5.65
Q5_K_M	2.78 GB	5.50
Q5_K_S	2.74 GB	5.21
Q5_0	2.75 GB	5.21
Q4_1	2.52 GB	4.78
Q4_K_M	2.40 GB	4.58
Q4_K_S	2.33 GB	4.37
IQ4_NL	2.32 GB	4.50
Q4_0	2.32 GB	4.34
IQ4_XS	2.21 GB	4.25
Q3_K_L	2.15 GB	4.03
Q3_K_M	2.02 GB	3.74
IQ3_M	1.94 GB	3.66
Q3_K_S	1.86 GB	3.41
IQ3_S	1.87 GB	3.44
IQ3_XS	1.80 GB	3.30
IQ3_XXS	1.66 GB	3.06
Q2_K	1.64 GB	2.96
Q2_K_S	1.56 GB	2.96
IQ2_M	1.55 GB	2.70
IQ2_S	1.48 GB	2.50
TQ2_0	1.42 GB	2.06
IQ2_XS	1.36 GB	2.31
IQ2_XXS	1.27 GB	2.06
TQ1_0	1.28 GB	1.69
IQ1_M	1.18 GB	1.75
IQ1_S	1.13 GB	1.56
Q1_0	0.87 GB	1.125

Quality Benchmarks

Quality measured with MMLU-Pro (128 random questions across 14 categories, 5-shot CoT) and MuSR (128 multi-step reasoning questions). CoT reasoning enabled server-side via --reasoning on. Recovery = accuracy relative to bf16 baseline.

Results filling in as evaluation progresses.

Quant	Size	MMLU-Pro	MuSR
bf16	7.77 GB	81.2%	pending
Q8_0	4.13 GB	81.2%	40.6%
Q6_K	3.19 GB	pending	pending
Q5_1	2.95 GB	pending	pending
Q5_K_M	2.78 GB	pending	pending
Q5_K_S	2.74 GB	pending	pending
Q5_0	2.75 GB	pending	pending
Q4_1	2.52 GB	pending	pending
Q4_K_M	2.40 GB	pending	pending
Q4_K_S	2.33 GB	pending	pending
IQ4_NL	2.32 GB	pending	pending
Q4_0	2.32 GB	pending	pending
IQ4_XS	2.21 GB	pending	pending
Q3_K_L	2.15 GB	pending	pending
Q3_K_M	2.02 GB	pending	pending
IQ3_M	1.94 GB	pending	pending
Q3_K_S	1.86 GB	pending	pending
IQ3_S	1.87 GB	pending	pending
IQ3_XS	1.80 GB	pending	pending
IQ3_XXS	1.66 GB	pending	pending
Q2_K	1.64 GB	pending	pending
Q2_K_S	1.56 GB	pending	pending
IQ2_M	1.55 GB	pending	pending
IQ2_S	1.48 GB	pending	pending
TQ2_0	1.42 GB	pending	pending
IQ2_XS	1.36 GB	pending	pending
IQ2_XXS	1.27 GB	pending	pending
TQ1_0	1.28 GB	pending	pending
IQ1_M	1.18 GB	pending	pending
IQ1_S	1.13 GB	pending	pending
Q1_0	0.87 GB	pending	40.6%
Q2_0	1.60 GB	pending	pending

Perplexity Reference

On the calibration set (107 chunks, 4K context):

Quant	PPL	Δ vs bf16
bf16	11.64	—
Q8_0	11.71	+0.07
Q4_K_M	11.68	+0.04
Q3_K_M	12.57	+0.93
Q2_K	13.89	+2.24
IQ1_S	44.83	+33.19

Note: perplexity measures next-token prediction loss, not task accuracy. Q4_K_M is functionally near-lossless for this model.

Usage

llama.cpp Server

llama-server -m Nanbeige4.2-3B-Q4_K_M.gguf \
  --flash-attn on \
  --cache-type-k vbr \
  --batch-size 64 --ubatch-size 64 \
  --parallel 1 --port 8081

CLI

llama-cli -m Nanbeige4.2-3B-Q4_K_M.gguf \
  -p "Write a Python function that..." -n 512

Recommended Settings

Parameter	Value	Reason
`--flash-attn`	`on`	Force flash attention on for GPU memory efficiency (default: auto-detect)
`--cache-type-k`	`vbr`	Variable bit-rate KV cache — maps physical GPU pages on demand, enabling long context in limited VRAM
`--batch-size`	`64`	Lower batch size reduces peak VRAM with minimal throughput impact at this model scale (default: 2048)
`--parallel`	`1`	Single slot avoids cross-slot overhead at 3B scale (parallel >1 is exponentially slower)
`--reasoning`	`on`	Enable reasoning/thinking mode for multi-step reasoning

Hardware Requirements

Quant	Min RAM	Min VRAM (24K ctx)
Q6_K	5 GB	3.2 GB
Q4_K_M	4 GB	2.4 GB
Q3_K_M	4 GB	2.0 GB
Q2_K	3 GB	1.6 GB
Q1_0	2 GB	0.87 GB

References

License

Apache 2.0

Downloads last month: 3,277

GGUF

Model size

4B params

Architecture

nanbeige

Hardware compatibility

1-bit

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

16-bit

View +1 variant

Model tree for gkraker04/Nanbeige4.2-3B-GGUF

Base model

Nanbeige/Nanbeige4.2-3B-Base

Finetuned

Nanbeige/Nanbeige4.2-3B

Quantized

(39)

this model