惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

T
Tailwind CSS Blog
人人都是产品经理
人人都是产品经理
博客园 - 叶小钗
大猫的无限游戏
大猫的无限游戏
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
博客园 - 【当耐特】
The Cloudflare Blog
博客园 - 聂微东
博客园 - 司徒正美
量子位
博客园 - 三生石上(FineUI控件)
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
G
Google Developers Blog
Apple Machine Learning Research
Apple Machine Learning Research
罗磊的独立博客
酷 壳 – CoolShell
酷 壳 – CoolShell
Y
Y Combinator Blog
S
SegmentFault 最新的问题
T
The Blog of Author Tim Ferriss
P
Proofpoint News Feed
Google DeepMind News
Google DeepMind News
Blog — PlanetScale
Blog — PlanetScale
有赞技术团队
有赞技术团队
A
About on SuperTechFans

Ubergizmo

Are Robots Coming For Airport Jobs? Japan Airlines Testing Humanoids Apple Introduces Commitment-Based Monthly Subscriptions For The App Store No Google Drive Or iCloud! WhatsApp Reportedly Developing Independent Cloud Backup Infrastructure OnePlus Ace 6 Ultra Launches With 8,600mAh Battery And 165Hz Gaming Display Adobe Launches Firefly AI Assistant To Automate Workflows Across Photoshop And Premiere The AI Did Not Go Rogue. The Company Gave It Too Much Access Leica Announces M-A Hammertone Limited Edition For 20th Anniversary Vivo Y600 Pro Announced With Massive 10,200 mAh Battery And 1.5K OLED Display AI Was Supposed To Cut Costs. Now Some Companies Say It Costs More Than Workers Will AI Data Centers Make Power Bills Worse? A Massive Utah Project Shows Why People Are Worried Google Workspace Redesign: 13 New App Icons Feature Modern Gradient Look MediaTek Dimensity 7450 And 7450X Announced With 200MP Support And AI for Mid-Range Phones Microsoft Looks Ready To Stop Forcing Windows Updates At The Worst Time Tesla Cybercab Production Starts For Elon Musk’s Autonomous Robotaxi BYD Yangwang U8L Dingzang Edition: Ultra-Luxury SUV With Zero-Gravity Seats XChat Standalone Messaging App Launches For X Users On iPhone And iPad Pokémon Scarlet And Violet Players Can Download A Free Chesnaught Raid Event Until April 30 Goodbye, Windows! SteamOS 3.9 Brings Official Support For ASUS ROG Ally, Lenovo Legion Go, And AMD Handhelds Geely Eva Cab Robotaxi, An AI-Powered Autonomous EV With 4ms Reaction Time New Honor MagicPad 3 Pro 12.3 Debuts Bringing 165Hz OLED Screen And Snapdragon 8 Gen 5 LPDDR6 RAM: Faster, Smarter Memory For The Next Generation Of Tech OpenAI Introduces Workspace Agents: Autonomous Collaboration For ChatGPT Teams Xbox Cut Game Pass Prices But Ended Day-One Call of Duty Access Govee Launches $99.99 Solar String Lights for Smart Backyard Lighting Insta360 Updates Flow 2 Series Gimbals With Enhanced Android Support And Apple Integration Microsoft Integrates Advanced “Agent Mode” With AI Into Word, Excel And PowerPoint Volkswagen Jetta Concept X Unveiled: A Strategic Electric Shift Under 15,000 USD DJI Unveils New Entry-Level Lito Drone Series: 48 MP, 4K Video, ActiveTrack And More Fitbit Enhances Sleep Score With Deep Analytics And Digital Coaching Threads Introduces Live Chats for Real-Time Event Interaction
Microsoft Launches New AI To Rival Google Nano Banana In ...
Paulo Montenegro · 2026-06-05 · via Ubergizmo

Microsoft has introduced its latest artificial intelligence model, MAI-Image-2.5, aiming to challenge the current industry benchmark for realistic image generation, Google’s Nano Banana 2. Announced on Tuesday, June 2, during the Build 2026 developer conference, the new technology enters a highly competitive digital media landscape.

The technology is deployed in two distinct versions designed to fulfill different operational requirements. The standard MAI-Image-2.5 framework is engineered to edit and generate visual assets with elevated precision and user control. Concurrently, Microsoft introduced MAI-Image-Flash, a variant specifically optimized for processing efficiency and speed.

Data from the independent AI Arena benchmarks, which were highlighted during the official Build presentation, indicate that Microsoft’s MAI-Image-2.5 outperforms Google’s Nano Banana 2 specifically in image editing capabilities. However, evaluation metrics confirm that the model still trails OpenAI’s GPT-Image-2 in editing performance, though its competitive positioning remains significant within the market.

Official demonstrations published on Microsoft’s corporate website illustrate the model’s capacity to modify images without introducing common digital artifacts. While this clean generation process benefits creative professionals seeking seamless edits, technical analysts note it simultaneously increases the difficulty of identifying AI-generated deepfakes.

The new model has been integrated into Microsoft’s software ecosystem for immediate evaluation. Enterprise users can access the technology within the PowerPoint application and through Microsoft Foundry, the company’s dedicated marketplace for corporate AI solutions. Additionally, both the high-precision MAI-Image-2.5 and the efficiency-focused MAI-Image-Flash variants are available for public testing at no cost through Microsoft’s testing platform.

Key Model Variants

  • MAI-Image-2.5: Developed for high-precision generation and detailed editing control.
  • MAI-Image-Flash: Streamlined version optimized for maximum computational efficiency.

Platform Availability

  • Microsoft PowerPoint: Integrated directly for productivity workflows.
  • Microsoft Foundry: Accessible via the corporate AI marketplace.
  • Web Portal: Available for external public testing without charge.

Filed in . Read more about and .