Microsoft launched seven MAI models at Ignite, including image and real-time voice tools with preview pricing.

Microsoft unveils seven MAI models at Ignite

Microsoft unveils seven MAI models at Ignite

Image: AI-generated editorial illustration

Microsoft used Ignite 2026 to unveil a seven-model MAI lineup aimed at coding, cybersecurity, image generation and editing, speech, transcription and automated decision-making. The company said the launch, announced on Nov. 17, was designed for customers facing rising token costs as they run agents continuously, and it paired the reveal with preview pricing for its new image and real-time voice models.

Microsoft announced seven new MAI models at its Ignite 2026 event on Nov. 17, introducing tools for coding, cybersecurity, image generation and editing, speech, transcription and automated decision-making. The lineup includes MAI Cyber 2 Flash, MAI Code 2, MAI Code 2 Flash, MAI Image 2.7, MAI Realtime 1, Dragon Copilot Transcribe and Decision 1.1.

The company framed the release around customer concerns about rising token costs, especially for developers and enterprises that run agents continuously. In that context, Microsoft said the new models are meant to provide more scalable price-performance for business use cases, with preview availability and usage-based pricing on the newest image and voice products.

What did Microsoft launch at Ignite?

Microsoft launched a seven-model MAI lineup built for enterprise workflows, with each model targeting a different part of the product stack. The company said the set spans coding, cybersecurity, image generation and editing, speech, transcription and decision-making, and it presented the launch as part of Ignite 2026 at San Francisco’s Chase Center.

The models named in the announcement were MAI Cyber 2 Flash, MAI Code 2, MAI Code 2 Flash, MAI Image 2.7, MAI Realtime 1, Dragon Copilot Transcribe and Decision 1.1. Microsoft said it released the lineup on Nov. 17, 2026.

Independent reporting from Interesting Engineering said the models were aimed at enterprise use cases including coding, cybersecurity, image editing, speech, transcription and automated decision-making. FN News likewise said Microsoft unveiled the models at Ignite 2026 in San Francisco.

How is Microsoft pricing MAI Realtime 1 and MAI Image 2.7?

Microsoft said MAI Realtime 1 starts at $0.01 per audio minute for base speech-to-speech usage, with reasoning billed separately at $1 per million input tokens and $5 per million output tokens. The company described it as its first speech-to-speech model for low-latency conversational applications and said it is intended for enterprise conversations.

Microsoft said MAI Realtime 1 outperformed GPT Realtime 2.1 on quality, speed and price across public benchmarks and Microsoft product evaluations, while noting that comparisons vary by application and methodology. The company said the model sits on the performance-price Pareto frontier for Big Bench Audio and Tau-Voice benchmarks.

For image work, Microsoft announced MAI Image 2.7 and MAI Image 2.7 Flash in public preview through Microsoft Foundry at about $35 and $10 per 1,000 images, respectively. Microsoft said the flagship model supports image generation and editing, while Flash is aimed at lower-cost, higher-volume production workloads focused on speed and efficiency.

Microsoft said both image models sit on the quality-cost Pareto frontier for generation and editing. It also said MAI Image 2.7 ranked first on Artificial Analysis image editing and that Arena ranked MAI as a top-two lab in both image generation and editing. The company described the Flash version as a lower-cost model for production-ready visuals.

What do the image models add for builders and enterprises?

Microsoft says MAI Image 2.7 can create and edit images from text, source images, web context and visual references. On the model page, the company said users can attach up to five reference images, and that those images are uploaded only on submission and after any required consent is accepted.

The company said MAI Image 2.7 is intended for both casual and professional use cases, while Microsoft Labs described the image tool as a limited preview for personal, non-commercial use only and said it should not be used for school or work-related activities. Microsoft also said Canvas is powered by the Flash model for creating, editing and refining images.

Microsoft said the broader seven-model launch is designed to address rising token costs for developers and enterprises running agents continuously, and it positioned the models as a lower-cost option for repeat generation and editing tasks.

What else was included in the MAI lineup?

Beyond image and voice, Microsoft’s Ignite post said Dragon Copilot Transcribe is built on MAI Transcribe 2 and already powers 10 million patient encounters a month. Microsoft said the transcription system has halved multilingual error rates.

Decision 1.1 is the company’s latest decision model, and Microsoft said it is 50 times faster and 79 times cheaper than GPT-Sol 6.1. Microsoft described that comparison as a company-reported benchmark and pricing claim. Independent reporting also said Decision 1.1 is used by 4,000 customers.

Microsoft said Maia 200, its second-generation AI accelerator, is now in production running its largest models and achieves 1.7x more tokens per watt. The company’s Ignite announcement said the seven-model launch fits its broader effort to deliver more price-efficient AI infrastructure and models for enterprise deployments.

Sources