Microsoft Revolutionizes AI with New Multimodal Models

Microsoft Expands Its AI Horizon with Multimodal Models

Microsoft has recently ramped up its efforts in the field of artificial intelligence by unveiling three new models that go well beyond simple language models. In an announcement made on Thursday, the company introduced these innovations, which include a voice and text transcription model, as well as a new version of its image model. These developments mark a significant milestone in Microsoft's strategy to diversify its AI capabilities.

An Innovative Voice Transcription Model

Among these new offerings, the voice transcription model stands out as a first for Microsoft. This model is capable of converting audio recordings into text in no less than 25 different languages. Designed for various applications such as video subtitling, meeting transcription, and voice assistance, this model can handle audio recordings of up to 60 seconds in length. Microsoft claims that this technology is a major asset for businesses looking to enhance their communication and accessibility.

Improvements in Image Generation

In parallel, Microsoft has also introduced the second generation of its internal image model. This model promises not only faster generation speeds but also more realistic visual representations. These improvements aim to surpass the previous model, thus providing users with more powerful tools for creating visual content. The new models are already available on Microsoft's Foundry and MAI playground platforms, and there are plans to integrate MAI-Image-2 into popular services like Bing and PowerPoint.

A Diversification Strategy for Microsoft

These developments illustrate Microsoft's commitment to expanding its offerings in the competitive AI market. Microsoft's Copilot chatbot, already well-established among businesses using Office 360 and Azure, is an example of this strategy. By focusing on secure and enterprise-friendly text models, Microsoft aims to differentiate itself from its competitors. The new AI tools, such as Copilot Cowork and Copilot Health, reflect this ambition.

Microsoft Facing Competition

As a major player in technology, Microsoft has the resources necessary to invest in ambitious projects that even some well-funded startups, like OpenAI, cannot always afford. Recently, OpenAI announced the discontinuation of its video application Sora AI to refocus on its core activities. In a context where the AI industry in 2026 aims to demonstrate the utility of its tools in professional settings, Microsoft appears well-positioned to meet this challenge.

The Challenge of Generative Media

Generative media, which includes the creation of images and videos by AI, requires considerable computational and energy resources. Microsoft, like Google, is investing heavily in research to make these technologies more efficient and less energy-intensive. Google recently announced that it would continue to explore generative media while seeking to optimize the profitability and energy efficiency of its models, such as the new video model Veo 3.1 Lite.