字号·· | 护眼
罗塞塔简报

微软AI发布聚焦降本的企业级模型阵容及授权权重

点「原文对照」整页切到原文,或双击某段只看那段的原文。

。2026年11月18日,微软AI宣布推出一系列新的企业级AI模型和MAI授权权重,旨在降低部署成本并保障客户数据控制权。

Microsoft AI Unveils Cost-Focused Enterprise Model Lineup and Licensed Weights On November 18, 2026, Microsoft AI announced a family of new enterprise AI models and MAI Licensed Weights designed to lower deployment costs and preserve customer data control.

阵容包括用于安全监控的MAI Cyber 2 Flash、GitHub Copilot中的MAI Code 2和Code 2 Flash、Microsoft Foundry中的MAI Image 2.7和Image 2.7 Flash、用于语音转语音交互的MAI Realtime 1、用于医疗文档的Dragon Copilot Transcribe,以及Decision 1.1模型路由器。

The lineup includes MAI Cyber 2 Flash for security monitoring, MAI Code 2 and Code 2 Flash in GitHub Copilot, MAI Image 2.7 and Image 2.7 Flash in Microsoft Foundry, MAI Realtime 1 for speech-to-speech interaction, Dragon Copilot Transcribe for medical documentation, and the Decision 1.1 model router.

微软AI还推出了MAI授权权重,使企业能够在其自身Azure环境中使用专有数据定制前沿模型。

Microsoft AI also introduced MAI Licensed Weights, enabling enterprises to customize frontier models with proprietary data inside their own Azure environments.

公司报告称,在与竞争对手模型的对比中实现了激进的成本降低和具有竞争力的基准测试分数,同时还包括一项AT&T定制5G分析部署。

The company reported aggressive cost reductions and competitive benchmark scores against rival models, alongside an AT&T custom 5G analytics deployment.

阅读媒体原文 ↗