Instructions to use ubergarm/Qwen3.6-27B-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 ubergarm/Qwen3.6-27B-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 ubergarm/Qwen3.6-27B-GGUF:IQ4_NL
# Run inference directly in the terminal:
llama cli -hf ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ubergarm/Qwen3.6-27B-GGUF:IQ4_NL
# Run inference directly in the terminal:
llama cli -hf ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

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 ubergarm/Qwen3.6-27B-GGUF:IQ4_NL
# Run inference directly in the terminal:
./llama-cli -hf ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

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 ubergarm/Qwen3.6-27B-GGUF:IQ4_NL
# Run inference directly in the terminal:
./build/bin/llama-cli -hf ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

Use Docker

docker model run hf.co/ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

LM Studio
Jan

vLLM

How to use ubergarm/Qwen3.6-27B-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ubergarm/Qwen3.6-27B-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": "ubergarm/Qwen3.6-27B-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

Ollama
How to use ubergarm/Qwen3.6-27B-GGUF with Ollama:
```
ollama run hf.co/ubergarm/Qwen3.6-27B-GGUF:IQ4_NL
```
Unsloth Desktop

How to use ubergarm/Qwen3.6-27B-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

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": "ubergarm/Qwen3.6-27B-GGUF:IQ4_NL"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use ubergarm/Qwen3.6-27B-GGUF with Docker Model Runner:
```
docker model run hf.co/ubergarm/Qwen3.6-27B-GGUF:IQ4_NL
```

Lemonade

How to use ubergarm/Qwen3.6-27B-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

Run and chat with the model

lemonade run user.Qwen3.6-27B-GGUF-IQ4_NL

List all available models

lemonade list

Hermes Agent

How to use ubergarm/Qwen3.6-27B-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 ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

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 ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use ubergarm/Qwen3.6-27B-GGUF with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf ubergarm/Qwen3.6-27B-GGUF:IQ4_NL

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 "ubergarm/Qwen3.6-27B-GGUF:IQ4_NL" \
  --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"

Testing IQ5_KS

by shewin - opened May 22

Discussion

shewin

May 22

Layer 64: 1288.28, 6192.00, 7480.28 MiB (output layer)

Total : 16703.73, 8997.88, 25701.60 MiB
Memory required for model tensors + cache: 26990 MiB
Memory available on all devices - compute: 28656 MiB
llm_load_tensors: ggml ctx size = 0.86 MiB
llm_load_tensors: offloading 64 repeating layers to GPU
llm_load_tensors: offloading non-repeating layers to GPU
llm_load_tensors: offloaded 65/65 layers to GPU
llm_load_tensors: CPU buffer size = 985.16 MiB
llm_load_tensors: CUDA0 buffer size = 17992.04 MiB
...........................................................................................
llama_init_from_model: n_ctx = 80128
llama_init_from_model: n_batch = 8192
llama_init_from_model: n_ubatch = 8192
llama_init_from_model: flash_attn = 1
llama_init_from_model: attn_max_b = 4096
llama_init_from_model: fused_moe = 1
llama_init_from_model: grouped er = 1
llama_init_from_model: fused_up_gate = 1
llama_init_from_model: fused_mmad = 1
llama_init_from_model: rope_cache = 0
llama_init_from_model: graph_reuse = 1
llama_init_from_model: k_cache_hadam = 0
llama_init_from_model: v_cache_hadam = 0
llama_init_from_model: split_mode_graph_scheduling = 0
llama_init_from_model: reduce_type = f16
llama_init_from_model: sched_async = 0
llama_init_from_model: ser = -1, 0
llama_init_from_model: freq_base = 10000000.0
llama_init_from_model: freq_scale = 1
llama_kv_cache_init: CUDA0 KV buffer size = 2810.13 MiB
llama_init_from_model: KV self size = 2660.50 MiB, K (q8_0): 1330.25 MiB, V (q8_0): 1330.25 MiB
llama_init_from_model: CUDA_Host output buffer size = 0.95 MiB
llama_init_from_model: CUDA0 compute buffer size = 8080.00 MiB
llama_init_from_model: CUDA_Host compute buffer size = 1412.22 MiB
llama_init_from_model: graph nodes = 2981
llama_init_from_model: graph splits = 2
llama_init_from_model: enabling only_active_experts scheduling

main: n_kv_max = 80128, n_batch = 8192, n_ubatch = 8192, flash_attn = 1, n_gpu_layers = 99, n_threads = 101, n_threads_batch = 101

PP	TG	N_KV	T_PP s	S_PP t/s	T_TG s	S_TG t/s
8192	2048	0	2.408	3401.56	31.581	64.85
8192	2048	8192	2.505	3270.63	33.048	61.97
8192	2048	16384	2.789	2936.80	35.122	58.31
8192	2048	24576	2.804	2921.27	36.345	56.35
8192	2048	32768	2.967	2760.67	38.090	53.77
8192	2048	40960	3.136	2612.40	39.949	51.27

shewin

May 22

Not top-tier but good quality and very fast

ubergarm

Owner May 24

Yes, I'm using the IQ4_KS with MTP on my 24GB VRAM 3090 GPU for daily use. It is fast and good enough to get some things done easier than doing it myself. Probably the first local model that is "good enough" for me haha...

For this dense model, probably not much benefit going with large batch sizes (the MoE benefit a lot from that usually). Using MTP is going to speed up decode a lot too for coding style workloads.

I love all your wild demo screenshots!

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment