Microsoft unveils new AI models to lessen reliance on OpenAI and lower costs for developers

In a bold move to enhance its position in the artificial intelligence landscape, Microsoft has unveiled new models aimed at reducing dependency on third-party providers like OpenAI. During the Build developer conference held in San Francisco, the tech giant introduced MAI-Code-1, its first model designed to transform written descriptions into application and website source code. The rise of AI coding, often referred to as vibe coding, allows both seasoned developers and those without programming expertise to generate complex software using simple text prompts. This trend has prompted Microsoft to create its own models, which can be operated on its Azure cloud infrastructure, thereby circumventing the costs associated with external services. Microsoft's announcement comes on the heels of Google's release of the Gemini 3.5 Flash model, capable of coding and executing various tasks. Alongside MAI-Code-1, the company is also rolling out MAI-Thinking, a reasoning model engineered for high efficiency and performance at a lower token cost, as highlighted by Kyle Daigle, Microsoft's head of developer marketing. Tokens are essentially the currency developers use to access these AI models, making cost efficiency a crucial factor. As Microsoft seeks to expand its footprint within the AI stack, it faces competition from rapidly growing firms like OpenAI and Anthropic, the latter of which recently filed for an IPO. Microsoft has significantly invested in both companies, pouring in $13 billion into OpenAI and $5 billion into Anthropic. Currently, MAI-Thinking-1 is available for private preview through Microsoft Foundry, allowing selected customers to test the model before its public release. The MAI-Code-1 model is integrated into the GitHub Copilot AI coding service and the Visual Studio Code text editor, marketed as "inference ultra-efficient." Additionally, Microsoft has introduced updated cloud-based solutions for speech recognition, synthetic voice generation, and image generation, along with small Aion models that can be executed on Windows PCs.