Instructions to use ZeroXClem/Qwen3-4B-MiniMight with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ZeroXClem/Qwen3-4B-MiniMight with Transformers:

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

pipe = pipeline("text-generation", model="ZeroXClem/Qwen3-4B-MiniMight")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ZeroXClem/Qwen3-4B-MiniMight")
model = AutoModelForCausalLM.from_pretrained("ZeroXClem/Qwen3-4B-MiniMight", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use ZeroXClem/Qwen3-4B-MiniMight with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ZeroXClem/Qwen3-4B-MiniMight"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ZeroXClem/Qwen3-4B-MiniMight",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ZeroXClem/Qwen3-4B-MiniMight

SGLang

How to use ZeroXClem/Qwen3-4B-MiniMight 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 "ZeroXClem/Qwen3-4B-MiniMight" \
    --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": "ZeroXClem/Qwen3-4B-MiniMight",
		"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 "ZeroXClem/Qwen3-4B-MiniMight" \
        --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": "ZeroXClem/Qwen3-4B-MiniMight",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ZeroXClem/Qwen3-4B-MiniMight with Docker Model Runner:
```
docker model run hf.co/ZeroXClem/Qwen3-4B-MiniMight
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

🧠 ZeroXClem/Qwen3-4B-MiniMight

"Small in scale. Mighty in mind."
A beautifully blended 4B model with 262k context length, fusing deep reasoning, code, safety, and creativity — distilled through MergeKit’s model_stock magic.

🔧 Merge Configuration

name: ZeroXClem/Qwen3-4B-MiniMight
base_model: Qwen/Qwen3-4B-Thinking-2507
dtype: bfloat16
merge_method: model_stock
models:
  - model: TeichAI/Qwen3-4B-Thinking-2507-Claude-4.5-Opus-High-Reasoning-Distill
  - model: bunnycore/Qwen3-4B-Mini-Merge
  - model: bunnycore/Qwen3-4B-MegaMerge
  - model: bunnycore/Qwen3-4B-Max-Merge
  - model: ertghiu256/qwen3-4b-claude-sonnet-x-gemini-reasoning
  - model: ZeroXClem/Qwen3-4B-Hermes-Axion-Pro
tokenizer_source: Qwen/Qwen3-4B-Thinking-2507

🌟 Highlights

🧠 Reasoning-Centric Intelligence — Optimized for multi-step thought, explanation, STEM logic, and symbolic problem solving.
🪐 262,144 Token Context Window — Massive memory span for long documents, complex chains, or multi-turn workflows.
🧬 Deep Merge of Code, RP, and Claude/Gemini Logic — Captures both fluency and fidelity in multi-domain generation.
🔐 Safe Alignment from Hermes & Axion — Includes red-teamed safety-conscious merges to guide outputs.
✍️ Elegant Dialogue & Creative Thought — Engages in immersive roleplay, structured writing, and storytelling.
⚡ Efficient 4B Inference — Small enough for local GPUs, quantized formats, and mobile deployments.

🧪 Usage Example

from transformers import AutoTokenizer, AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("ZeroXClem/Qwen3-4B-MiniMight", device_map="auto", torch_dtype="auto")
tokenizer = AutoTokenizer.from_pretrained("ZeroXClem/Qwen3-4B-MiniMight")

messages = [{"role": "user", "content": "Explain quantum tunneling with an example."}]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=384)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

🧠 Use enable_thinking=True or /think with compatible frontends for enhanced reasoning control.

🔁 Models Merged

Source Model	Specialization
`TeichAI/Qwen3-4B-Thinking-2507-Claude-4.5-Opus-High-Reasoning-Distill`	High reasoning distilled from Claude Opus
`bunnycore/Qwen3-4B-Mini-Merge`	Mini-coder, Qwaifu, Darwin blend
`bunnycore/Qwen3-4B-MegaMerge`	AesCoder + OSS GPT + Qwaifu fusion
`bunnycore/Qwen3-4B-Max-Merge`	Maximal blend with strong instruction coherence
`ertghiu256/qwen3-4b-claude-sonnet-x-gemini-reasoning`	Gemini 3 Pro & Claude Sonnet enhanced
`ZeroXClem/Qwen3-4B-Hermes-Axion-Pro`	Reasoning + alignment + safety-core logic

💼 Ideal For

🔍 Chain-of-thought reasoning & symbolic logic
🧪 Long-document summarization & analysis
💬 Roleplay & storytelling with high memory
📚 Tutoring & educational AI (math, code, science)
⚖️ Safety-aligned assistants for deployment
👨‍💻 Code generation, refactoring, and walkthroughs

🚫 Limitations

Long responses may truncate if max_new_tokens is too low
Mixed training sources may result in style variance across domains
Dialogue tone defaults to helpful/instructive rather than emotional/creative

📜 License

Apache 2.0 Please verify license alignment when adapting for commercial or public-facing use.

💌 Credits

Thank you to bunnycore, Qwen Team, ertghiu256, TeichAI for the wonderful fine tunes and base models that made this merge possible! Built by ZeroXClem Team — merging distillations, hugging intellect and ideas into tangible weights. MergeKit mastery allows full distilled insight in one mighty little package.

Mini in name. Infinite in mind. 🛸

Downloads last month: 15

Safetensors

Model size

4B params

Tensor type

BF16

Model tree for ZeroXClem/Qwen3-4B-MiniMight

Qwen/Qwen3-4B-Thinking-2507

TeichAI/Qwen3-4B-Thinking-2507-Claude-4.5-Opus-High-Reasoning-Distill

ZeroXClem/Qwen3-4B-Hermes-Axion-Pro

bunnycore/Qwen3-4B-Max-Merge

bunnycore/Qwen3-4B-Mini-Merge

ertghiu256/qwen3-4b-claude-sonnet-x-gemini-reasoning

Merge model

this model

Quantizations

5 models