OpenAI Rival Ahead? Microsoft Reveals Two In-House AI Models For The First Time

Microsoft has introduced two new in-house AI models which seems to rival OpenAI and other AI giants in the future. Here is what are these AI models are all about.

Edited By: Divya | Published By: Divya | Published: Aug 29, 2025, 01:37 PM (IST)

Microsoft In house AI models: What's coming?

techlusive.in Written By article news — Written By Divya

Microsoft has always openly supported one of the well-known AI giants, OpenAI. The wider user base and constant updates of ChatGPT, even benefited OpenAI over time. However, it seems like Microsoft is now also focusing on its own in-house AI models. Mustafa Suleyman, CEO of Microsoft AI, has revealed in a post on X that the tech giant is testing two in-house AI models — MAI-Voice-1 and MAI-1-preview.Also Read: OpenAI Confirms Adult-Only ChatGPT With Custom Personalities And Erotic Conversations

If you are thinking about Microsoft’s Copilot, then know that it is primarily dependent on OpenAI’s GPT tech. With the upcoming in-house AI models, it seems like Microsoft is trying to reduce its dependence on OpenAI and emerge as an independent AI giant.Also Read: Gemini Nano Banana AI Image Editor Rolls Out To Google Search And NotebookLM: Here’s What’s New

MAI-Voice-1 and MAI-1-preview: All About Microsoft’s In-House AI Models

Microsoft AI (MAI) has previewed two in-house AI models:Also Read: How To Create Unique Rangoli Designs For Diwali Using Gemini Nano Banana AI Tool: Check Prompt

MAI-Voice-1 which is a natural speech generation model which runs on a single GPU. It can generate a full minute of audio in under a second on a single GPU. Microsoft claim it to be “one of the most efficient speech systems available today.”

Where to try: If you are excited to try out the upcoming AI models of Microsoft, then you can try out the MAI-Voice-1 in Copilot and Copilot Labs, where you can even try the expressive speech and storytelling demos.

On the other hand, the MAI-1-preview is a text-based foundation model. It is trained on around 15,000 NVIDIA H100 GPUs. It will essentially be like ChatGPT, helping with everyday queries.

Where to try: It is available for public testing on LMArena for community model evaluation. Apart from this, Microsoft revealed that it will be available for certain text use cases within Copilot in the coming weeks to get feedback.