Instructions to use kaitchup/LFM3-1.7B-SLQ-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 kaitchup/LFM3-1.7B-SLQ-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 kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K
# Run inference directly in the terminal:
llama cli -hf kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K
# Run inference directly in the terminal:
llama cli -hf kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K

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 kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K
# Run inference directly in the terminal:
./build/bin/llama cli -hf kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K

Use Docker

docker model run hf.co/kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K

Llama new
LM Studio
Jan

vLLM

How to use kaitchup/LFM3-1.7B-SLQ-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "kaitchup/LFM3-1.7B-SLQ-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": "kaitchup/LFM3-1.7B-SLQ-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K

Ollama
How to use kaitchup/LFM3-1.7B-SLQ-GGUF with Ollama:
```
ollama run hf.co/kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K
```
Unsloth Desktop

How to use kaitchup/LFM3-1.7B-SLQ-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K

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": "kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use kaitchup/LFM3-1.7B-SLQ-GGUF with Docker Model Runner:
```
docker model run hf.co/kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K
```

Lemonade

How to use kaitchup/LFM3-1.7B-SLQ-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K

Run and chat with the model

lemonade run user.LFM3-1.7B-SLQ-GGUF-Q6_K

List all available models

lemonade list

Hermes Agent

How to use kaitchup/LFM3-1.7B-SLQ-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 kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K

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 kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use kaitchup/LFM3-1.7B-SLQ-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K

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 "kaitchup/LFM3-1.7B-SLQ-GGUF:Q6_K" \
  --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"

Subscribe to my newsletter · Support

LFM3-1.7B-SLQ-GGUF

This is LiquidAI/LFM3-1.7B, compressed with a customized version of SLQ (Statistically-Lossless Quantization) and assembled as GGUF. The search combines existing quantized donor weights and chooses a source separately for 154 complete matrices, including the shared embeddings and output head.

The search used many donors, but could only match the size of AtomicChat's AD-Q6_K with a slightly better EAR@10: 97.7383% versus 97.5171%, an improvement of 0.2212 percentage points. Both GGUF files are approximately 1.588 GB.

One model is available:

Model	GGUF file	Active text weights (GB)	Complete file (GB)	Measured EAR@10 ↑
EAR97.7	LFM3-1.7B-SLQ-EAR97.7-Q6_K.gguf	1.579	1.588	97.738262%

Measured EAR@10: 97.738262%, evaluated using the BF16 reference's top 10 tokens.
Sizes: decimal GB (1 GB = 1,000,000,000 bytes). Active text weights include fixed text tensors and exclude metadata and alignment.
Embeddings / output head: shared Q8_0 weights.
The file contains 287 text tensors. No MTP tensors, vision tensors or multimodal projector are included.
Made by: The Kaitchup.
License: LFM Open License v1.0.

The filename suffix Q6_K is a Hub compatibility and dominant-precision-family label for this custom SLQ allocation.

Method: SLQ paper · SLQ repository.

Candidate sources

The search considered Q4_K_M, Q5_K_M, Q6_K, Q8_0, QAD-Q4_0, and BF16 from LiquidAI/LFM3-1.7B-GGUF, plus AD-IQ3_S, AD-IQ4_XS, AD-Q4_K_M, AD-Q5_K_M, AD-Q6_K, and Q8_0 from AtomicChat/LFM3-1.7B-GGUF. Identical tensor payload alternatives were canonicalized.

99% of the work was already done thanks to the GGUFs released by LiquidAI and AtomicChat, and the SLQ implementation by ISTA.

Credit goes to Liquid AI for the base model and official GGUF donors, AtomicChat for its GGUF donors, and Michael Helcig, Eldar Kurtic and Dan Alistarh for SLQ. The base and donor repositories declare LFM Open License v1.0 licensing; the full license is included in LICENSE. Original copyright and attribution rights remain with their authors.

Precision by layer

EAR97.7

SVG

EAR comparison

EAR means Expected Acceptance Rate. It measures the overlap between the original BF16 model's next-token probabilities and those of the compared model. For EAR@10, both distributions are separately renormalized over the BF16 reference's 10 most likely tokens. An EAR@10 of 0.99 means 99% average probability overlap on that measured support; it does not mean 99% benchmark accuracy or identical generated answers.

Zoomable SVG

All points use the same 512 × 2,048-token corpus and BF16 reference. Donor measurements used four parallel sequences; the SLQ checkpoint used eight.

SLQ's “statistically lossless” objective concerns approximate fidelity, not identical numerical outputs. This release achieves the measured EAR@10 shown above. The same calibration corpus guided model selection, so these values are calibration fidelity rather than an independent held-out benchmark. Statistical equivalence on downstream tasks has not been established for this release.

Instructions: llama.cpp

Use a llama.cpp build with LFM3 architecture support. Native text scoring used an NVIDIA GeForce RTX 3060 with llama.cpp commit 2973c945bf32759976d858a3a35ae0e026b4507d. The exported GGUF's tensor formats, allocation and exact file size were checked against the measured model.

Install the Hugging Face CLI, then download one file:

pip install -U huggingface_hub
hf download kaitchup/LFM3-1.7B-SLQ-GGUF \
  LFM3-1.7B-SLQ-EAR97.7-Q6_K.gguf --local-dir .

Start an OpenAI-compatible text server with the downloaded model:

llama-server \
  --model LFM3-1.7B-SLQ-EAR97.7-Q6_K.gguf \
  --n-gpu-layers 99 \
  --ctx-size 8192 \
  --port 8080

Send a chat request:

curl http://localhost:8080/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{
    "messages": [{"role": "user", "content": "Is Earth smaller than Venus?"}],
    "max_tokens": 256
  }'