Microsoft Unveils New AI Models to Challenge OpenAI and Anthropic

Microsoft Unveils New AI Models to Challenge OpenAI and Anthropic
Nov 18, 2026 10:12
Nov 18, 2026 13:58

Microsoft has unveiled a series of new artificial intelligence (AI) models specialising in coding and cybersecurity as it seeks to gain an edge over rivals such as Anthropic and OpenAI in the AI race. The announcement was made on Tuesday (November 17) at the Microsoft Ignite conference in San Francisco, United States.

At the conference, the company introduced MAI Code 2, a computer programming model. Microsoft claims that the model delivers capabilities comparable to those of leading competing AI models while costing significantly less to operate.

The company also unveiled a specialised cybersecurity model designed to identify and address software security flaws and vulnerabilities. It will initially be made available to a limited number of customers. In addition, Microsoft introduced new AI models for image generation and audio-to-text transcription.

High Performance at Lower Cost and a Focus on Security

In an interview with Reuters, Mustafa Suleyman, head of Microsoft's AI division, said the company's strategy of developing specialised technologies tailored to different types of tasks was helping it keep costs down. This approach, he said, allows Microsoft to match the capabilities of its competitors while, in some cases, achieving top rankings in performance benchmarks.

He said this was not simply about a single model or a temporary breakthrough, but rather an integrated system designed to maintain a balance between performance and cost.

Referring to the recent incident in which the Hugging Face platform was hacked through an autonomous agent developed by OpenAI, Suleyman said the latest cybersecurity risks had encouraged him to place greater emphasis on safety mechanisms rather than focusing solely on improving model capabilities.

He noted that his top priority was ensuring that AI does not take harmful actions and remains guided by human instructions—a concept known in research as ‘alignment’.

//DBTech/BMT/OR//