QClaw-4B-GGUF is the quantized GGUF version of LakoMoor/QClaw-4B — a 4-billion parameter model fine-tuned for agentic tasks and tool use, designed for use with OpenClaw-compatible agent frameworks.

This repository provides GGUF files for local inference with llama.cpp, Ollama, LM Studio, Jan, and other compatible runtimes.

Available Quantizations

Filename	Quant	Size	Quality	Recommended
`QClaw-4B-F16.gguf`	F16	~8 GB	Maximum	Servers / high VRAM
`QClaw-4B-Q8_0.gguf`	Q8_0	~4.5 GB	Excellent	High quality inference
`QClaw-4B-Q5_K_M.gguf`	Q5_K_M	~3 GB	Very good	⭐ Best balance
`QClaw-4B-Q4_K_M.gguf`	Q4_K_M	~2.5 GB	Good	⭐ Most popular
`QClaw-4B-Q3_K_M.gguf`	Q3_K_M	~2 GB	Medium	Low RAM devices

Usage

llama.cpp

bash

./llama-server \
  -m QClaw-4B-Q4_K_M.gguf \
  -a qclaw-4b \
  --jinja \
  --port 8000

Ollama

ollama run hf.co/LakoMoor/QClaw-4B-GGUF:Q4_K_M

LM Studio

Search for LakoMoor/QClaw-4B-GGUF in the model browser and select your preferred quantization.

Model Details

Base model: LakoMoor/QClaw-4B
Architecture: Decoder-only transformer (Qwen3.5-4B based)
Parameters: ~4B
Quantization tool: llama.cpp
Primary use case: Agentic workflows, tool calling, multi-step reasoning

Intended Use

QClaw-4B-GGUF is intended for:

Local inference on consumer hardware (CPU and GPU)
Agentic pipelines using OpenClaw or compatible frameworks
Tool-augmented assistants requiring compact, efficient inference
Research into small-model agent capabilities

Out-of-scope use: Not intended for safety-critical systems without additional alignment work.

Training annotation cards and dataset curation provided by Aleksandr Nikolich — Love. Death. Transformers..

Citation

bibtex

@misc{qclaw4b2026,
  title  = {QClaw-4B: State-of-the-Art 4B Agent Model for OpenClaw},
  author = {Nikolay Kompanets (LakoMoor)},
  year   = {2026},
  url    = {https://huggingface.co/LakoMoor/QClaw-4B}
}

License

Apache 2.0