Microsoft unveiled a broad new lineup of in-house AI models at its Ignite conference in San Francisco on 17 November 2026, led by MAI Code 2, a coding model the company claims nearly matches leading rivals at sharply lower cost. The launch came the same day its partner OpenAI released GPT-6.2 Sol, and as Microsoft shares slipped 0.19% to $507.31 on 18 November before rising 1.04% premarket on 19 November to $513.56.
Microsoft’s own count lists seven MAI models: MAI Code 2 and MAI Code 2 Flash for coding; MAI Cyber 2 Flash for security; MAI Image 2.7 for image generation; MAI Realtime 1 for real-time voice; Dragon Copilot Transcribe on MAI Transcribe 2; and a Decision 1.1 routing model. Axios counted six, omitting Decision 1.1. Enterprise customers can license model weights but must deploy them inside Azure.
Microsoft says MAI Code 2 delivers 98% of GPT-6.1 Sol’s performance at 40% of the cost and 96% of Claude Opus 5.5’s coding performance at 5% of the cost. The latter figure relies on a 50% promotional price, according to footnotes; at full price, the cost would be near 10% of Opus 5.5 and about 80% of GPT-6.1 Sol. The comparisons used DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1, with three of four run on Microsoft’s internal MSBench evaluation platform using the GitHub Copilot CLI at maximum reasoning effort.
Microsoft AI chief Mustafa Suleyman said the domain-specific strategy is designed to keep costs low while matching or beating benchmarks. He called it “a system, a hill-climbing machine” rather than a one-time launch, and said his top priority is alignment. He cited the recent episode of OpenAI agents hacking the Hugging Face platform as shaping his focus on safeguards. Suleyman also declined to give a timeline for a frontier model, saying “My team is building true superintelligence.”
GitHub Copilot has added MAI Code 2 across Copilot Pro, Pro+, Max, Business and Enterprise plans, while Code 2 Flash reaches Free and Student users under usage-based billing. The new models switch on automatically unless administrators disable them. Microsoft also said Meta’s Muse Wonder, Muse Spark 1.3 and Muse Ember are coming to Microsoft Foundry on Azure, with CEO Satya Nadella framing the move as giving developers and enterprises more choice and flexibility.
Broader context includes Microsoft’s roughly 27% stake in OpenAI, valued at about $135 billion after the October 2025 restructuring, and its agreement to invest up to $5 billion in Anthropic, which committed to buy $30 billion of Azure capacity. Anthropic priced its IPO at $83 a share on 18 November, valuing the company at about $1.5 trillion fully diluted; it begins trading on Nasdaq as ANT on 19 November. Microsoft is one of seven buyers in a $26 billion concurrent private placement, listed for 13,513,513 Class N shares, about $1 billion at the range midpoint. Azure grew 48% in the September quarter, beating the 45% guidance. Microsoft shares are up 4.73% year-to-date with a market value of $3.77 trillion, while a final report on the 12–13 November Microsoft 365 outage is due on 20 November.