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

Libraries

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

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

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

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ZeroXClem/Qwen3-4B-Valiant-Polaris")
model = AutoModelForCausalLM.from_pretrained("ZeroXClem/Qwen3-4B-Valiant-Polaris")
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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

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

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ZeroXClem/Qwen3-4B-Valiant-Polaris"
# 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-Valiant-Polaris",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

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

SGLang

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

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

🧠 ZeroXClem/Qwen-4B-Valiant-Polaris

Overview

ZeroXClem/Qwen-4B-Valiant-Polaris is a thoughtfully blended model crafted using Model Stock merging via MergeKit. It fuses the structured reasoning of Polaris, the creative expressiveness of Dot-Goat and RP-V3, and the scientific depth of ShiningValiant3 into a powerful 4B architecture built atop the official Qwen/Qwen3-4B.

Designed for enhanced reasoning, uncensored creativity, deep roleplay, and advanced agentic performance — this model is both lightweight and intellectually formidable.

🔧 Merge Details

Merge Method: model_stock
Base Model: Qwen/Qwen3-4B
Dtype: bfloat16
int8_mask: true
normalize: false
Tokenizer Source: Qwen/Qwen3-4B

Merge Configuration

models:
  - model: bunnycore/Qwen3-4B-Dot-Goat
  - model: bunnycore/Qwen3-4B-RP-V3
  - model: POLARIS-Project/Polaris-4B-Preview
  - model: ValiantLabs/Qwen3-4B-ShiningValiant3
  - model: Qwen/Qwen3-4B
merge_method: model_stock
base_model: Qwen/Qwen3-4B
normalize: false
int8_mask: true
dtype: bfloat16
tokenizer_source: Qwen/Qwen3-4B

🧬 Models Merged

🐐 `bunnycore/Qwen3-4B-Dot-Goat`

Uncensored, multi-domain, LoRA-infused Qwen model focusing on creativity, tool-use, and deep chat alignment.

🎭 `bunnycore/Qwen3-4B-RP-V3`

Character-rich roleplay personality fusion from the amoral, mixture-of-thought, and SuperbEmphasis trees.

🌌 `POLARIS-Project/Polaris-4B-Preview`

Post-trained with advanced reinforcement learning on reasoning-heavy datasets. Surpasses Claude Opus & Grok on math and logic benchmarks.

✨ `ValiantLabs/Qwen3-4B-ShiningValiant3`

Expertly aligned to scientific reasoning, agentic workflows, and multi-domain creative logic.

🔧 `Qwen/Qwen3-4B`

Official pretrained Qwen3 model with support for thinking / non-thinking modes, multilingual reasoning, and tool-calling capabilities.

✨ Features & Highlights

🔹 Advanced Reasoning — Polaris post-training brings SOTA performance in chain-of-thought, math, and symbolic logic.

🔹 Roleplay & Uncensored Expressiveness — RP-V3 and Dot-Goat contribute dynamic personas and emotion-rich conversational modeling.

🔹 Scientific & Engineering Alignment — ShiningValiant3 ensures excellent handling of complex scientific and analytical queries.

🔹 Multimodal-Friendly & Tool-Aware — Qwen’s native agentic design enables external tool use and seamless task execution.

🔹 Lightweight Excellence — At just 4B parameters, this model performs impressively for its size with long context (32k+) and efficient inference.

🎯 Use Cases

💬 Conversational & RP Agents
📚 Scientific Reasoning & Educational Tutoring
🔍 Advanced Math & Logic Problem Solving
✍️ Creative Writing & Storyworld Simulation
🧠 Tool-Integrated Autonomous Agents

🚀 Usage Instructions

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "ZeroXClem/Qwen-4B-Valiant-Polaris"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)

prompt = "Solve: What is the smallest prime greater than 100?"
messages = [{"role": "user", "content": prompt}]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=True
)

inputs = tokenizer([text], return_tensors="pt").to(model.device)
outputs = model.generate(**inputs, max_new_tokens=256)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

🧭 Alignment & Ethics

⚠️ Unfiltered Behavior: Some sub-models are uncensored and may produce unmoderated content. Please implement safety layers when deploying in public-facing apps.

⚠️ Responsible Use: Outputs are governed by their inputs. Always review critical output for bias, hallucination, or ethical misalignment.

📜 License: Apache 2.0 + governed by respective base model licenses (see individual repos).

💌 Feedback & Contributions

Got thoughts, benchmarks, or new merge suggestions? We’d love to hear from you! Feel free to:

Submit issues or pull requests 💡
Tag us in your Hugging Face projects ❤️
Join the discussion around merging and alignment at @ZeroXClem on HF and GitHub!

ZeroXClem Team | 2025 ✨