Instructions to use ubergarm/Qwen3-30B-A3B-Thinking-2507-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/Qwen3-30B-A3B-Thinking-2507-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/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K
# Run inference directly in the terminal:
llama cli -hf ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K
# Run inference directly in the terminal:
llama cli -hf ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K

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/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K
# Run inference directly in the terminal:
./llama-cli -hf ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K

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/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K
# Run inference directly in the terminal:
./build/bin/llama-cli -hf ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K

Use Docker

docker model run hf.co/ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K

LM Studio
Jan

vLLM

How to use ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ubergarm/Qwen3-30B-A3B-Thinking-2507-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/Qwen3-30B-A3B-Thinking-2507-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K

Ollama
How to use ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF with Ollama:
```
ollama run hf.co/ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K
```
Unsloth Desktop

How to use ubergarm/Qwen3-30B-A3B-Thinking-2507-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/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_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": "ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF with Docker Model Runner:
```
docker model run hf.co/ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K
```

Lemonade

How to use ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K

Run and chat with the model

lemonade run user.Qwen3-30B-A3B-Thinking-2507-GGUF-Q2_K

List all available models

lemonade list

Hermes Agent

How to use ubergarm/Qwen3-30B-A3B-Thinking-2507-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/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_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 ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_K

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use ubergarm/Qwen3-30B-A3B-Thinking-2507-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/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_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 "ubergarm/Qwen3-30B-A3B-Thinking-2507-GGUF:Q2_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"

ubergarm commited on Jul 30, 2025

Commit

e20c306

1 Parent(s): cc1c0ab

add perplexity graphs

Browse files

Files changed (2) hide show

README.md +13 -11
images/perplexity.png +3 -0

README.md CHANGED Viewed

@@ -31,14 +31,16 @@ Perplexity computed against *wiki.test.raw*.
 ![Perplexity Chart](images/perplexity.png "Chart showing Perplexity improving as BPW increases.")
-These first two are just test quants for baseline perplexity comparison:
 * `bf16` 56.894 GiB (16.007 BPW)
-  - Final estimate: PPL = TODO
 * `Q8_0` 30.247 GiB (8.510 BPW)
-  - Final estimate: PPL = TODO
 ## `IQ5_K` 21.324 GiB (5.999 BPW)
-Final estimate: PPL = TODO
 <details>
@@ -90,7 +92,7 @@ custom=$(
 </details>
 ## `IQ4_K` 17.878 GiB (5.030 BPW)
-Final estimate: PPL = TODO
 <details>
@@ -142,7 +144,7 @@ custom=$(
 </details>
 ## `IQ4_KSS` 15.531 GiB (4.370 BPW)
-Final estimate: PPL = TODO
 <details>
@@ -195,7 +197,7 @@ custom=$(
 ## `IQ3_K` 14.509 GiB (4.082 BPW)
-Final estimate: PPL = TODO
 <details>
@@ -247,7 +249,7 @@ custom=$(
 </details>
 ## `IQ3_KS` 13.633 GiB (3.836 BPW)
-Final estimate: PPL = TODO
 <details>
@@ -299,7 +301,7 @@ custom=$(
 </details>
 ## `IQ2_KL` 11.516 GiB (3.240 BPW)
-Final estimate: PPL = TODO
 <details>
@@ -351,7 +353,7 @@ custom=$(
 </details>
 ## `IQ2_KT` 9.469 GiB (2.664 BPW)
-Final estimate: PPL = TODO
 <details>
@@ -402,7 +404,7 @@ custom=$(
 </summary>
 ## `IQ1_KT` 7.583 GiB (2.133 BPW)
-Final estimate: PPL = TODO
 <details>

 ![Perplexity Chart](images/perplexity.png "Chart showing Perplexity improving as BPW increases.")
+These first three are just test quants for baseline perplexity comparison:
 * `bf16` 56.894 GiB (16.007 BPW)
+  - Final estimate: PPL = 7.3149 +/- 0.05076
 * `Q8_0` 30.247 GiB (8.510 BPW)
+  - Final estimate: PPL = 7.3284 +/- 0.05091
+* `Q4_0` 16.111 GiB (4.533 BPW)
+  - Final estimate: PPL = 7.4534 +/- 0.05151
 ## `IQ5_K` 21.324 GiB (5.999 BPW)
+Final estimate: PPL = 7.3440 +/- 0.05091
 <details>
 </details>
 ## `IQ4_K` 17.878 GiB (5.030 BPW)
+Final estimate: PPL = 7.3634 +/- 0.05104
 <details>
 </details>
 ## `IQ4_KSS` 15.531 GiB (4.370 BPW)
+Final estimate: PPL = 7.3861 +/- 0.05128
 <details>
 ## `IQ3_K` 14.509 GiB (4.082 BPW)
+Final estimate: PPL = 7.4360 +/- 0.05162
 <details>
 </details>
 ## `IQ3_KS` 13.633 GiB (3.836 BPW)
+Final estimate: PPL = 7.4959 +/- 0.05204
 <details>
 </details>
 ## `IQ2_KL` 11.516 GiB (3.240 BPW)
+Final estimate: PPL = 7.6992 +/- 0.05345
 <details>
 </details>
 ## `IQ2_KT` 9.469 GiB (2.664 BPW)
+Final estimate: PPL = 8.0207 +/- 0.05638
 <details>
 </summary>
 ## `IQ1_KT` 7.583 GiB (2.133 BPW)
+Final estimate: PPL = 8.8341 +/- 0.06231
 <details>

images/perplexity.png ADDED Viewed

Git LFS Details

SHA256: c18305a8cc7817f50f6c11088f81703247d838ab95926772f0a772b7a691e815
Pointer size: 131 Bytes
Size of remote file: 169 kB