Tag
Microsoft pressured LG to disable unwanted McAfee pop-up ads that were automatically installed on Windows 11 when connecting LG monitors, following user backlash.
Microsoft announces public preview of MAI-Image-2.5-Pro and MAI-Voice-2-Flash, their latest purpose-built generative AI models for image and voice, now available on Azure AI and powering Microsoft products like Bing Image Creator.
Microsoft confirms that LG monitors can automatically install the LG Monitor App Installer and McAfee ads via Windows Update without user consent, raising privacy concerns.
Microsoft launches the Xbox Backward Compatibility on PC program, making select original Xbox titles playable and purchasable on Windows PCs with enhanced graphical features and Game Pass integration.
Microsoft AI Frontiers open-sources MagenticLite, including the MagenticBrain 14B orchestrator and Fara 1.5 computer-use models, now available on Hugging Face with open weights and the full application stack.
Microsoft released Fara1.5-27B, a 27B parameter multimodal computer use agent for web browsers, fine-tuned from Qwen3.5-27B, capable of end-to-end web tasks via screenshots.
Microsoft is bringing original Xbox games to PC through backward compatibility, starting with four classic titles (Blinx, Conker, Crimson Skies, Fuzion Frenzy) available now with enhanced graphics and features, as part of a broader effort to preserve and expand Xbox gaming on PC.
Microsoft Build 2026 shifts focus from better AI models to building an operating system for AI agents, repositioning its entire portfolio around agent-centric computing.
Microsoft releases Mage, a family of lightweight 4B-parameter multimodal models for visual understanding and generation, including Mage-VL for image/video understanding and Mage-Flow for text-to-image generation and editing, designed for research and deployment on modest hardware.
Microsoft is testing Moonshot's Kimi model inside Copilot, highlighting a shift where users care more about utility than the underlying AI model.
Microsoft and Mistral AI expand their partnership to integrate Mistral's frontier models across Azure, Microsoft Foundry, Copilot Studio, and Azure Local, aiming to provide more AI deployment choices for customers.
An article explaining how Windows carries a persistent identifier (GDID) that tracks users even behind a VPN, and how to disable the services that send it out.
Microsoft releases Mage-Flow-Edit-Turbo, a compact 4B-scale generative model for efficient text-to-image generation and instruction-based image editing, achieving state-of-the-art competitive quality through co-designed tokenizer and backbone.
Microsoft releases Mage-Flow, a compact 4B-parameter foundation model for efficient native-resolution text-to-image generation and instruction-based image editing, achieving competitive quality against much larger models.
Microsoft's strategy for self-evolving agent skills, training them like neural networks with epochs, batch size, learning rates, and validation gates, fully open-source.
AutoGen, a framework for building multi-agent AI applications, is now in maintenance mode with no new features planned; Microsoft recommends its Agent Framework instead.
Microsoft open-sourced Resource2Skill, which automatically distills executable agent skills from human resources such as tutorial videos, articles, and code, enabling real-world applications like web pages, PPTs, and Excel.
Mage-Flow is a compact 4B-parameter generative stack for efficient text-to-image generation and instruction-based image editing, featuring a co-designed lightweight tokenizer (Mage-VAE) and a native-resolution multimodal diffusion transformer trained with rectified flow matching. It achieves competitive performance while enabling high-resolution generation at 0.59s on a single A100 GPU.
AMD launches Helios, its first rack-scale AI system to rival Nvidia's offerings, with Microsoft joining as a customer. The system is set to ship later this year.
Microsoft open-sourced SkillOpt, a tool that enables AI agent skills to self-improve by automatically evaluating and rewriting their instructions, outperforming hand-crafted prompts and other optimizers with no model lock-in.