started · updated
Ant Digital Technology releases LingDT-Decision-1.0 decision model
Ant Digital Technology has launched LingDT-Decision-1.0, a new decision-oriented model designed for high-frequency enterprise tasks such as intelligent customer service, security risk control, and financial applications. Unlike general large language models (LLMs) that focus on text generation, this model is optimized for structured decision-making tasks, including intent recognition, task classification, tool selection, and risk assessment.
The model is available in 4B and 27B parameter versions. In testing across 23 open-source benchmarks, the 27B version outperformed several prominent models, including Microsoft-Decision-1 and Jev 1.13.0. Ant Digital Technology reports that in specific business scenarios, the model can reduce token consumption by over 90% and increase task processing speeds by up to 10 times compared to general LLMs.
A key innovation of LingDT-Decision-1.0 is its native support for multi-choice and sequential decision-making using Directed Acyclic Graph (DAG) task dependency representations. This allows the model to handle complex, multi-step business requests, such as identifying multiple user intents within a single query and determining the correct order of operations. The model is currently available on the Ant Digital Technology DTMaaS platform.