Ant Digital has launched LingDT-Decision-1.0, a lightweight model specifically designed for high-frequency structured decision-making tasks. Available in 4B and 27B parameter versions, the model introduces native multi-choice and sequential decision (DAG) capabilities, aiming to separate judgment-based tasks like intent recognition from general-purpose LLMs to optimize cost and latency.
Cross-verified reports indicate that in real-world scenarios such as intelligent customer service and risk control, LingDT-Decision-1.0 reduces token consumption by over 90% and improves response speed by approximately 10 times compared to traditional LLMs. In 23 open-source benchmarks, its 27B version achieved higher overall accuracy than competitors like Microsoft-Decision-1, establishing a performance lead at its parameter scale.
This release signifies a deepening shift in AI applications from 'general generation' to specialized 'decision-making.' By significantly lowering inference costs and enhancing real-time performance, the model offers developers a more cost-effective solution, facilitating scalable deployment on the DTMaaS platform and redefining efficiency standards for enterprise AI.