| ## Important: No commercial use without commercial license (yet) | |
| **Deca 3 Alpha Ultra: Model Overview** | |
| Deca 3 Alpha Ultra is a large language model built on a DynAMoE (Dynamically Activated Mixture of Experts) architecture, offering a distinct approach from traditional Mixture of Experts (MoE) models. With 4.6 trillion parameters, it leverages novel techniques to push the limits of natural language understanding and generation. | |
| ### Key Specifications | |
| * **Architecture:** DynAMoE (Dynamically Activated Mixture of Experts), designed for optimal performance. Derived from existing architectures | |
| * **Parameters:** 4.6 trillion. | |
| ### License and Usage | |
| * **Current License:** Restricted access for select research institutions and strategic partners. This is to prevent the model from slowing down Deca 2.5 | |
| * **Future License:** A revised licensing model will expand access while maintaining responsible deployment. | |
| **About Deca** | |
| Founded as a small AI company in the U.S., Deca has rapidly expanded after receiving key support from **GenLabs**, advancing AI research and developing groundbreaking models like Deca 3 Alpha Ultra. | |