Skip to content
VietVance

Site search

Science & Technology

Microsoft Unveils AI Models for Programming and Cybersecurity

Microsoft unveiled specialized AI models in San Francisco on November 17 for programming, cybersecurity, image generation and speech transcription. The company says its MAI Code 2 programming model approaches competing products’ performance at a lower price in tests it selected; its cybersecurity model will initially be available to a limited group of customers.

VietVance

1 min read

Microsoft unveiled a series of specialized artificial intelligence models focused on programming and cybersecurity at its Microsoft Ignite conference in San Francisco, United States, on November 17. It also introduced new models for image generation and speech transcription.

A Microsoft logo in New York, United States.(VietVance)

Among them, MAI Code 2 is designed primarily for programming. Microsoft says that, in performance tests it selected and published, the model achieved 98% of the programming quality of OpenAI’s GPT-6.1 Sol model at about 40% of its price. Compared with Anthropic’s Claude Opus 5.5, it achieved 96% of that model’s performance at about 5% of its cost. These comparisons are based on tests selected by Microsoft.

Microsoft also introduced a specialized cybersecurity model that it plans to offer initially to a limited group of customers to help identify and address software vulnerabilities. The company said AI tools are making security weaknesses easier to discover quickly.

Mustafa Suleyman, head of Microsoft’s AI division, said developing models for specific fields helps control costs while achieving performance close to or better than some competing products in benchmark tests. He said Microsoft is building a system of models that continually improves, with both performance and pricing optimized. On safety, he said preventing AI from carrying out harmful actions and ensuring it operates as intended are priorities in model development.

Source: VietVance

Related