US Edition
Microsoft Releases New AI Models to Expand Further Beyond OpenAI
The models are Microsoft’s first major in-house release since a March reorganization, as the company pushes to reduce reliance on OpenAI.
- On Thursday, Microsoft released three in-house AI models—MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2—exclusively on its Foundry developer platform.
- Developing these internal tools represents a strategic shift by Microsoft to reduce reliance on OpenAI, as CEO of Microsoft AI Mustafa Suleyman previously stated the goal is to make the company "self-sufficient in AI."
- Suleyman claimed MAI-Transcribe-1 is "the most accurate transcription model in the world," while MAI-Image-2 ranks in the top three on the Arena.ai leaderboard and integrates into Bing and PowerPoint.
- Enterprise customers now access these tools on Foundry, placing Microsoft in direct competition with OpenAI's Whisper and DALL-E, with Suleyman telling The Verge the transcription model runs at half the GPU cost of state-of-the-art alternatives.
- Microsoft continues to bolster its superintelligence team, recently hiring former Allen Institute CEO Ali Farhadi to support development, with plans to build frontier models to be "completely independent" if needed.
16 Articles
16 Articles
Microsoft released 3 new AI models, ramping up competition with its close partner, OpenAI
Zamek/VIEWpressMicrosoft has made three new in-house-built AI models available on its Foundry platform.It's a sign of the company reducing its reliance on its longtime partner, OpenAI.A new deal agreed in October allowed Microsoft more independence from OpenAI.Microsoft has released three AI models created in-house, in a move that signals a push toward greater independence from its longtime partner, OpenAI.MAI-Transcribe-1, MAI-Voice-1, and MAI-…
Times of India
The Economic TimesMicrosoft launches 3 AI models for transcription, image, and speech generation
Through these three models — MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2 — Microsoft aims to expand its push into multimodal AI capabilities for developers. The models are also being integrated into Microsoft products, including Copilot, Bing, and PowerPoint, with enterprise adoption already underway
Microsoft releases new AI models to expand further beyond OpenAI
Microsoft announced MAI-Transcribe-1, a new speech-to-text model, and made its in-house MAI-Voice-1 and MAI-Image-2 models broadly available to developers for commercial use for the first time, expanding its proprietary AI capabilities beyond its OpenAI partnership.
Microsoft AI has announced the launch of three own foundational models that generate text, voice and images, in what is the clearest signal to date that the company is building its own multimodal AI stack to reduce its reliance on OpenAI. MAI-Transcribe-1 transcribes voice to text in 25 languages and is 2.5 times faster than Microsoft’s current Azure Fast offer. MAI-Voice-1 generates audio from text and can produce 60 seconds of audio in a secon…
Microsoft is further expanding its portfolio of in-house AI models. Under the label "Microsoft AI," or MAI for short, three new models are now available via Microsoft Foundry – one for speech recognition, one for speech synthesis, and one for image generation. For those interested in AI-powered speech... Read the article: Microsoft expands AI portfolio: New models for speech and image available Where to follow us: Facebook, Reddit, Google News,…
Microsoft Releases Three Multimodal AI Models As It Expands In-House Capabilities
Microsoft has introduced three new foundational AI models capable of generating text, voice, and images, marking a step in its effort to expand its in-house multimodal AI capabilities while maintaining its partnership with OpenAI. The models, developed by the company’s Microsoft AI research division, aim to compete with offerings from other AI labs and position Microsoft with a broader, self-built model stack. Three Models Target Text, Voice, An…
Coverage Details
Bias Distribution
- 34% of the sources lean Left, 33% of the sources are Center, 33% of the sources lean Right
Factuality
To view factuality data please Upgrade to Premium
