Introducing Arcee-Nova

What a week here at Arcee AI. On the heels of Arcee-Scribe yesterday, today we bring you Arcee-Nova – our highest-performing open source model... Evaluated on the same stack as the OpenLLM Leaderboard 2.0, making it the top-performing open source model tested on that stack. Its performance approaches that of GPT-4 from May 2023, marking a significant milestone.

Nova is a merge of Qwen2-72B-Instruct with a custom model tuned on a generalist dataset mixture.

Performance

• Evaluated on the OpenLLM Leaderboard 2.0 stack
• Top-performing open source model on this stack
• Approaches GPT-4 (May 2023) performance levels

Technical Details

• Merge of Qwen2-72B-Instruct with a custom-tuned model, with RLHF on top of that
• Custom generalist dataset mixture used for tuning
• GGUF versions available on Hugging Face

Key Capabilities

1. Reasoning
2. Creative Writing
3. Coding
4. General Language Understanding

Business Applications

• Customer Service: Advanced chatbots and virtual assistants
• Content Creation: High-quality marketing and documentation
• Software Development: Code generation and quality improvement
• Data Analysis: Enhanced interpretation and reporting
• Research and Development: Literature reviews and hypothesis generation
• Legal and Compliance: Contract analysis and regulatory checks
• Education and Training: Adaptive learning systems

Performance Metrics

Acknowledgments

We thank the open source AI community for their ongoing contributions and the Qwen team for their foundational work on Qwen2-72B.

Looking Ahead

We invite researchers, developers, and businesses to explore Arcee-Nova's capabilities. Our commitment to open source AI advancement continues, and we look forward to seeing how the community builds upon this technology.

This post was created with assistance from Arcee-Nova.