Microsoft CEO Mustafa Suleyman told Axios that the company is making significant progress toward building its own frontier general-purpose AI model, as it introduced six specialized models designed to rival leading offerings from OpenAI and Anthropic at a fraction of the cost.
The six new models were announced at Microsoft’s Ignite conference in San Francisco. They cover coding, cybersecurity, image generation and voice.
The lineup includes MAI Code 2, a coding model that Microsoft says delivers 98% of GPT-6.1 Sol’s performance at 40% of the price and 96% of Anthropic’s Claude Opus 5.5’s performance at just 5% of the cost. Another model, MAI Code 2 Flash, is said to match GPT-6 Luna at 65% of the price and Claude Haiku 5.5 at 20% of the cost. Meanwhile, MAI Cyber 2 Flash, a cybersecurity model, reportedly outperforms GPT-6 Astra on a security benchmark at less than 2% of the cost per task.
READ: Microsoft employee with ‘200/200’ rating weighs exit package amid layoff fears (October 28, 2026)
Other models include MAI Image 2.7, which Microsoft says ranks first in image editing, along with new transcription and real-time voice models designed to compete with leading rivals.
The launch is significant because Microsoft has long relied on OpenAI and other AI companies for its most advanced models. The company’s strategy is to build models tailored to specific tasks, allowing them to deliver comparable performance at significantly lower costs.
“It basically means that you cram more tokens and knowledge and expertise into the same amount of parameters,” Suleyman told Axios. However, a model optimized for everyday coding tasks may be less capable when handling advanced mathematics or scientific research.
Microsoft’s performance comparisons rely on selected benchmarks, and some pricing advantages reflect promotional discounts. MAI Code 2 will be available in GitHub Copilot, potentially shifting some coding workloads away from external models.
Microsoft is also developing a frontier general-purpose large language model that can compete across a broader range of tasks. Suleyman said the company had achieved state-of-the-art pre-training results and was making strong progress in reinforcement learning, but declined to provide a timeline for releasing the model. He said his team was “building superintelligence.”
The new model could significantly change Microsoft’s relationship with external AI providers such as OpenAI.
READ: Microsoft offers buyouts to 60,000 workers as it shifts spending toward AI (October 26, 2026)
Microsoft is also offering enterprise customers access to licensed model weights, allowing them to customize its models using proprietary data while requiring deployment within Azure. Suleyman said the approach would “feel like open source,” while offering greater security and control.
Last week, core Microsoft 365 services, including Outlook, Copilot, Exchange Online and Teams, experienced a major global outage affecting businesses in 74 countries. Department of Homeland Security (DHS) Secretary Markwayne Mullin said artificial intelligence may have played a role in the high-profile technology disruptions in the U.S. over the past week.


