---
datasets:
- prithivMLmods/Open-Omega-Explora-2.5M
license: apache-2.0
language:
- en
base_model:
- Qwen/Qwen3-0.6B
pipeline_tag: text-generation
library_name: transformers
tags:
- text-generation-inference
- moe
- code
- science
- biology
- chemistry
- thinking
---

![1.png](/static-proxy?url=https%3A%2F%2Fcdn-uploads.huggingface.co%2Fproduction%2Fuploads%2F65bb837dbfb878f46c77de4c%2FA0T_etoZYm1NP6iae7nSW.png)

# **Explora-0.6B**

> **Explora-0.6B** is a lightweight and efficient **general-purpose reasoning model**, fine-tuned on **Qwen3-0.6B** using the first 100,000 entries of the **Open-Omega-Explora-2.5M** dataset. It is tailored for **science and code**-focused reasoning tasks, combining symbolic clarity with fluent instruction-following, ideal for exploratory workflows in STEM domains.

> \[!note]
> GGUF: [https://huggingface.co/prithivMLmods/Explora-0.6B-GGUF](https://huggingface.co/prithivMLmods/Explora-0.6B-GGUF)

## **Key Features**

1. **General-Purpose STEM Reasoning**
   Fine-tuned for **code and science problems**, the model handles symbolic reasoning, basic computations, and structured logic with clarity and fluency.

2. **Built on Qwen3-0.6B**
   Leverages the multilingual and instruction-tuned capabilities of **Qwen3-0.6B**, making it well-suited for lightweight deployments with strong core reasoning ability.

3. **Open-Omega-Explora Dataset**
   Trained on the **first 100k entries** of the **Open-Omega-Explora-2.5M** dataset, which includes a diverse mix of problems from math, code, and science domains.

4. **Balanced Thinking Mode**
   Supports moderate reasoning depth while avoiding excessive hallucination—great for **step-by-step problem solving**, **function generation**, and **explanatory output**.

5. **Compact & Deployable**
   At just **0.6B parameters**, it’s ideal for **offline environments**, **low-resource inference setups**, and **educational tools** requiring fast, reliable logic.

6. **Output Flexibility**
   Capable of producing answers in **Markdown**, **Python**, **JSON**, or plain text depending on the task—suitable for both human readability and integration into pipelines.


## **Quickstart with Transformers**

```python
from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "prithivMLmods/Explora-0.6B"

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

prompt = "Explain Newton's second law of motion with a Python code example."

messages = [
    {"role": "system", "content": "You are a helpful science and code reasoning assistant."},
    {"role": "user", "content": prompt}
]

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

model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=256
)
generated_ids = [
    output_ids[len(input_ids):] for input_ids, output_ids in zip(model_inputs.input_ids, generated_ids)
]

response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)
```

## **Intended Use**

* Educational and lightweight research tools
* General science and programming help
* Low-resource STEM assistant for code labs or classrooms
* Fast-response agent for structured reasoning tasks

## **Limitations**

* Not optimized for deep multi-hop reasoning or creative tasks
* May require prompt engineering for highly specific technical queries
* Smaller context window and lower fluency compared to larger models
* Best used with **specific and scoped questions** for accurate outputs