←News detail

English text is being prepared.Updated Oct 24 22:55

Ant Digital Technologies Releases Decision Model LingDT-Decision-1.0, Cutting Token Consumption by Over 90%

English text is being prepared.

AI summary

Ant Digital Technologies has released its decision model LingDT-Decision-1.0, purpose-built for rapid industrial decision-making, offering 4B and 27B versions. It is now live on the Ant Digital Technologies DTMaaS platform with a limited-time free trial. The model natively supports multiple-choice and sequential decision-making (DAG), and outperforms decision models such as Microsoft-Decision-1 and Jev 1.13 across 23 open-source evaluation benchmarks, with the 27B version achieving the highest accuracy among models of the same parameter size. Tests by Ant Digital Technologies in customer service, risk control, finance and other scenarios show that, when handling the same tasks, it can reduce Token consumption by over 90% compared with large language models and improve response time by 10x.

Why read this

AI score 62

The original text provides Token and response data for the decision model in real business scenarios, which can be used to assess the feasibility of replacing general-purpose large models with specialized small models for high-frequency decision-making.

Topics

Tags

Body

蚂蚁数科发布决策模型 LingDT-Decision-1.0,将 Token 消耗降低超 90%

2026-10-22 18:12 IT之家 - 浩渺

感谢IT之家网友 仙人掌 的线索投递!

IT之家 10 月 22 日消息,蚂蚁数科今日发布决策模型 LingDT-Decision-1.0。该模型专为快速产业决策打造,在处理结构化决策任务时,具备准确率更高、响应更快、Token 成本更低等优势。

据介绍,相比于其他决策模型,LingDT-Decision-1.0 首创支持原生多选与按序决策能力,更适配企业真实应用场景。在 23 项开源评测基准中,LingDT-Decision-1.0 的性能超过了 Microsoft-Decision-1、Jev 1.13.等热门决策模型,LingDT-Decision-1.0 的 27B 版本在同参数模型中取得最高准确率。

蚂蚁数科团队在客服、风控、金融等真实业务场景测试显示,在处理同样任务时,LingDT-Decision-1.0 相比于大语言模型,可将 Token 消耗降低超 90%,响应时效提升 10 倍,并且性能表现更优。

在企业实际应用中,存在大量高频的决策类任务。例如,在客服场景中,当收到“这个订单已经三天了还没发货,请尽快处理。”的问询后,AI 需要快速进行一系列判断和决策:问题类型(退款、催单还是投诉?)、工单紧急度(是否需要优先处理)、归属部门(工单派到哪个部门)以及风险判断(是否需要人工接管)。

传统大语言模型在这类场景中存在明显短板。大语言模型通常需要逐字(Token)进行推理,再生成输出文本,链路长、成本高,也更容易出现错误,性能不够稳定。决策模型正是针对这些痛点而设计,它的目标不是生成文本,而是做出快速、准确的判断。因此,天然适合承担企业业务中的高频决策型任务,如意图判断、工具选择、规则分类和结果打分等。

相比其他决策模型,蚂蚁数科 LingDT-Decision-1.0 更加强调面向企业真实应用场景打造。除了判断、单选、分类、评级等基础能力外,LingDT-Decision-1.0 还首创推出原生多选与按序决策(DAG)功能,能够更好适配企业业务中复杂、多步骤的决策需求。

LingDT-Decision-1.0 提供 4B 和 27B 两个版本。目前,模型已经在蚂蚁数科 DTMaaS 平台上线。开发者可限时免费体验。

Source

English text is being prepared.