Nace AI Open-Sources Drex 1.5, a 9B Decision Model
Nace AI releases Drex 1.5, an open-source 9B decision model that scores all options in one forward pass and achieves 58.08 on Decision Index 0.3.1.
Stock photo for illustration only, not from the actual event
- Nace AI has open-sourced Drex 1.5, a 9B parameter decision model
- Evaluates probabilities for every option in a single forward pass
- Scores 58.08 on Decision Index 0.3.1 with 128K token context support
Nace AI has officially announced the open-source release of Drex 1.5, a specialized decision-making model featuring 9 billion parameters. Unlike traditional language models that focus primarily on text generation, this architecture is engineered specifically to evaluate and score various choices.
A key capability of Drex 1.5 is its ability to return a probability score for every available option within a single forward pass. This streamlined approach dramatically accelerates processing speeds and enhances efficiency when handling complex evaluation tasks.
Stock photo for illustration only, not from the actual event
In benchmark evaluations, Drex 1.5 secures a score of 58.08 on Decision Index 0.3.1. Furthermore, the model is capable of reading and processing up to 128,000 tokens of context, making it highly suitable for large-scale data analysis and intricate scenario planning.
Nace AI's focus on building a dedicated decision model rather than a general-purpose LLM highlights a growing industry shift toward specialized AI architectures. By outputting probabilities for all options in a single pass, it bypasses iterative generation steps, offering rapid backend solutions for recommendation engines, automated ranking, and strategic planning workflows.
Source: MarkTechPost
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment