Resonance-01: Ozone Research

Model Description

Resonance-01 is a 14 billion parameter language model developed by Ozone Research, built on the Qwen/Qwen2.5-14B-Instruct framework. It’s a causal language model optimized for advanced text generation and reasoning.
MAJOR HIGHLIGHT: CHAIN-OF-THOUGHT (CoT) REASONING
Resonance-01 thinks logically, breaking down complex problems into clear, step-by-step solutions—setting it apart from other models. This is Ozone Research’s most advanced release to date.

Join Our Community

https://discord.gg/ozone

Intended Uses & Limitations

Resonance-01 leverages its CoT reasoning for research, content creation, and enterprise applications. Key uses include:

Text generation with logical flow
Question answering with detailed, step-by-step explanations
Creative writing with coherent narratives
Technical analysis requiring structured problem-solving

Limitations:

May generate biased or inaccurate content; safeguards are recommended.
Performance varies by task and input complexity.

How to Use

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "ozone-research/Resonance-01"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)

prompt = "Explain why the sky is blue."
input_ids = tokenizer(prompt, return_tensors="pt").input_ids

generation_output = model.generate(input_ids, max_length=100)
print(tokenizer.decode(generation_output[0]))

Evaluation

Benchmarks

Benchmark data is pending, but Resonance-01’s CoT reasoning is expected to excel on MMLU Pro, particularly in problem-solving tasks. Full results coming soon.

Training Details

Training Infrastructure: 1x H100
Training Procedure: 3x Epoch using LoRA

Contact

For inquiries, reach out at info@ozone-ai.com or visit https://www.ozone-ai.com. You can also contact via discord at https://discord.gg/ozone

Attribution

Built with Qwen. Users must comply with the Qwen license agreement.

ozone-research
/

milo-1