Ant Digital Technologies, a subsidiary under the umbrella of Ant Group, has rolled out its innovative decision-making model, LingDT-Decision-1.0. This model is tailored specifically for swift industrial decision-making processes. When it comes to managing structured decision-making tasks, LingDT-Decision-1.0 stands out with its superior accuracy, rapid response times, and notably lower token costs. It is the pioneer in natively supporting both multiple-choice and sequential decision-making, aligning it more closely with real-world enterprise application scenarios.

Against a backdrop of 23 open-source evaluation benchmarks, LingDT-Decision-1.0's performance outshines that of several well-known decision-making models. Notably, its 27B version claims the highest accuracy among models boasting the same parameters. In real-world tests across customer service, risk control, finance, and various other business domains, the model slashes token consumption by over 90% and accelerates response speeds tenfold compared to traditional large language models, showcasing its exceptional performance.

Designed to tackle the challenges posed by lengthy chains, high costs, susceptibility to errors, and erratic performance in conventional large language models when dealing with high-frequency enterprise decision-making tasks, LingDT-Decision-1.0 goes beyond merely providing fundamental decision-making capabilities. It is engineered to better cater to the intricate, multi-step decision-making demands of enterprises. The model is offered in two variants: a 4B version and a 27B version, catering to diverse business needs.