metadata

base_model: unsloth/deepseek-r1-distill-qwen-32b-bn
tags:
  - text-generation-inference
  - transformers
  - unsloth
  - qwen2
  - trl
license: apache-2.0
language:
  - en
datasets:
  - Daemontatox/math_conv
library_name: transformers

Zirel-R1: Optimized for Fast and Essential Reasoning

Model Overview

Zirel-R1 is an advanced reasoning-optimized model designed for short, fast, and necessary reasoning, avoiding long and unnecessary computation. It surpasses Cogito-R1 and PathfinderAI S1 in efficiency, making it ideal for applications requiring structured logical inference and quick decision-making.

Developed by: Daemontatox
Model Series: Zirel
Base Model: unsloth/deepseek-r1-distill-qwen-32b
License: Apache-2.0
Languages: English
Finetuned on: Daemontatox/math_conv
Library: Transformers

Key Features

✅ Fast and Concise Reasoning – Delivers precise answers with minimal computational overhead.
✅ Optimized for Short-Form Problem Solving – Excels in extracting core insights efficiently.
✅ Enhanced Logical Inference – Ideal for applications in structured decision-making, math reasoning, and controlled AI workflows.

Usage

You can load the model using transformers:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "Daemontatox/Zirel-R1"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name, device_map="auto")

prompt = "What is the next number in the sequence: 2, 4, 8, 16?"
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
output = model.generate(**inputs, max_new_tokens=50)
print(tokenizer.decode(output[0], skip_special_tokens=True))

Performance

Speed: 🚀 Optimized for rapid inference and low-latency responses.

Accuracy: 🎯 Fine-tuned on high-quality mathematical and reasoning datasets.

Efficiency: ⚡ Processes only the necessary information for an answer.


Citation

If you use Zirel-R1, please cite:

@misc{daemontatox2025zirel,
  author = {Daemontatox},
  title = {Zirel-R1: Optimized for Fast and Essential Reasoning},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co./Daemontatox/Zirel-R1}
}

License

This model is released under the Apache-2.0 License.


---

This template ensures your model card looks professional and informative on Hugging Face. Let me know if you need modifications!