Instructions to use byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF", dtype="auto")

llama-cpp-python

How to use byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF",
	filename="Qwen3-30B-A3B-Instruct-2507-IQ3_S-2.69bpw.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use byteshape/Qwen3-30B-A3B-Instruct-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 byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S
# Run inference directly in the terminal:
llama cli -hf byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S
# Run inference directly in the terminal:
llama cli -hf byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

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 byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S
# Run inference directly in the terminal:
./llama-cli -hf byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

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 byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S
# Run inference directly in the terminal:
./build/bin/llama-cli -hf byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

Use Docker

docker model run hf.co/byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

LM Studio
Jan

vLLM

How to use byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

SGLang

How to use byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF with Ollama:
```
ollama run hf.co/byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S
```

Unsloth Studio

How to use byteshape/Qwen3-30B-A3B-Instruct-2507-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 byteshape/Qwen3-30B-A3B-Instruct-2507-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 byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF to start chatting

How to use byteshape/Qwen3-30B-A3B-Instruct-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 byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

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": "byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use byteshape/Qwen3-30B-A3B-Instruct-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 byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

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 byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

Run Hermes

hermes

Atomic Chat new

OpenClaw new

How to use byteshape/Qwen3-30B-A3B-Instruct-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 byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

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 "byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S" \
  --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 byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF with Docker Model Runner:
```
docker model run hf.co/byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S
```

Lemonade

How to use byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull byteshape/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_S

Run and chat with the model

lemonade run user.Qwen3-30B-A3B-Instruct-2507-GGUF-Q4_K_S

List all available models

lemonade list

Speed on RTX 5090

by islameissa - opened Feb 9

Discussion

islameissa

Feb 9

I wonder what are the setting that gets you 340 tps on the RTX 5090
I run the most recent build of llama.cpp and I get 230 tps (generation)
if the 340 is an average of token ingestion and generation then my settings are fine. But otherwise, if you meant 340 and more for token generation, I really want to know what are the settings that you are using.
Thanks

Ali93H

ByteShape org Feb 10

For those experiments, we used the default parameters on the b6974 release of llama.cpp (with CUDA 12.9), so they were not run on the most recent build.
Also, just to clarify the number: the fastest model we released reached 314 TPS for token generation, not 340. If you spotted 340 anywhere in our docs, please point us to it so we can double-check.
For performance reporting, we run each model 40 times using different prompts.
For this specific model (labeled as 2 in 5090 figure: Qwen3-30B-A3B-Instruct-2507-IQ3_S-2.75bpw.gguf), token generation TPS was:
• Average: 314.238
• Min: 305.459
• Max: 316.328

Happy to share more details if helpful.

islameissa

Feb 10

Sorry. That was my mistake. I looked at the graph without my glasses.
314 is still impressive. I use llama-server to spin the model. maybe it is slower because I'm on windows (assuming you are using Linux)
if you also use llama-server I would like to know the param like -c or if you use -ctk -ctv to reduce the vram ... etc. that will be very helpful to me.

Ali93H

ByteShape org Feb 10

No worries at all,
Yes, we’re running on Linux and using llama-server as well. For the performance benchmarks, we kept things simple and used the default llama.cpp settings for that build, without additional tuning flags.
From what I recall for that version, the defaults were roughly:
• Context length (-c): 4096
• KV cache type: f16 (no -ctk / -ctv overrides)

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