Microsoft Moves Beyond OpenAI With 7 New AI Models For Image Generation, Coding, Reasoning, More

Hyderabad: Microsoft AI (MAI), the company's core Artificial Intelligence (AI) division, announced seven new in-house developed AI models at the company's annual developer conference this year, Build 2026. These models include MAI-Thinking-1, MAI-Code-1-Flash, MAI-Image-2.5, MAI-Image-2.5 Flash, MAI Transcribe-1.5, MAI-Voice-2, and MAI-Voice-2-Flash, designed to handle individual specialised tasks ranging from reasoning, coding, image generation, transcription, and voice generation, respectively.

Microsoft mentioned that "Beyond these models, we’re building a superintelligence lab – a system and an approach we believe will define the next phase of AI." The Redmond-based tech giant noted that the processing power required "to train frontier models has increased by a factor of one trillion", which is expected to increase another thousand-fold over the next three years. The company mentions that this will provide advanced capabilities to the AI models and make the technology more effective.

MAI model family

The seven AI models together form the MAI model family, a multimodal ecosystem designed to handle real-world tasks across a range of modalities.

MAI-Thinking-1: It is Microsoft AI's flagship reasoning model. A medium-sized model trained from the ground up on clean data — without distillation from third-party models. It matches leading models on key software engineering benchmarks, demonstrates advanced mathematical reasoning, and is said to outperform Claude Sonnet 4.6 in blind human side-by-side evaluations.

MAI-Code-1-Flash: It is an inference-efficient agentic coding model built for and deeply integrated into GitHub Copilot, VS Code, and the broader Microsoft stack. With five billion active parameters, it is positioned as comparable to Claude Haiku but at a lower cost.

MAI-Image-2.5 and MAI-Image-2.5-Flash: They are image generation models that convert text to an image and also support image editing. It is worth noting that the Flash variant is the ultra-efficient and affordable version of the main image model. Microsoft claims that its image AI model surpasses the Arena score of Nano Banana Pro.

MAI-Transcribe-1.5: Microsoft says it is the world's leading transcription model, offering state-of-the-art accuracy at five times the speed of competing models, with built-in support for domain-specific terminology across 43 languages.

MAI-Voice-2 and MAI-Voice-2-Flash: Both AI models deliver natural-sounding speech generation across 15 languages, with the ability to adapt to a speaker's voice from a short audio sample. It includes safeguards against misuse, and the lower-cost Flash variant is described as coming soon.

Availability and Developer Access

In addition to distribution via Microsoft Foundry and integration into first-party Microsoft products, the MAI models will be available to developers through OpenRouter, Fireworks, and Baseten. Microsoft has also confirmed that developers will, for the first time, be able to fine-tune model weights directly.

All seven models share the same infrastructure and adhere to the same enterprise-grade data lineage standards. According to Microsoft, they are designed to work in concert and integrate directly into the products people use daily.