AI Aulis

AI NEWS IN ENGLISH · 18 November 2026

The most important AI news every morning. AI writes and checks the stories against the original sources – check important facts in the sources.

32 stories · next issue tomorrow morning · How the paper is made →

← Issue 18 November 2026Paper

Microsoft says MAI Code 2 reaches 98 per cent of GPT-6.1 Sol’s score at 40 per cent of the price

AI-generated contentWritten by AI

github.blog, axios.com, aitimes.com, x.com · 7 sources · Sources published: 17–18 November 2026

Microsoft’s new coding models, MAI Code 2 and MAI Code 2 Flash, can now be selected in GitHub Copilot. The performance figures are the company’s own, and the price comparison used a 50 per cent promotional price.

Microsoft unveiled six of its own AI models for coding, cybersecurity, images and speech at its Ignite event in San Francisco on Tuesday, Axios reported. The most important is the coding model MAI Code 2, which GitHub added to Copilot the same day.

MAI Code 2 is a mixture-of-experts (MoE) model with 2.8 trillion parameters, Korea’s AI Times reported. When the model runs, 80 billion of them are in use at a time. Its context window is 1.05 million tokens, and it also accepts images.

Mustafa Suleyman, CEO of Microsoft AI, told Axios that specialisation pays off because more knowledge and expertise fit into the same number of parameters. The downside, Axios wrote, is that a model optimised for everyday coding does not perform as well in demanding mathematics or scientific research.

Microsoft says MAI Code 2 reaches 98 per cent of GPT-6.1 Sol’s score, averaged across four coding benchmarks, at 40 per cent of the price. Against Anthropic’s Claude Opus 5.5, the figures are 96 per cent and 5 per cent. These numbers come from Microsoft itself. The cost comparison used a 50 per cent promotional price, AI Times reported, and Axios noted that the comparisons rest on selected benchmarks.

Nor is the comparison made against OpenAI’s latest model. OpenAI released GPT-6.2 Sol the same day. In early testing, it found nearly twice as many real issues in code review as GPT-6.1 Sol, GitHub said. Microsoft’s comparison does not include it.

The lighter MAI Code 2 Flash is aimed at routine fixes and small edits. Microsoft says it matches OpenAI’s GPT-6 Luna at 65 per cent of the price and Claude Haiku 5.5 at 20 per cent, Axios reported. However, Microsoft’s description as relayed by AI Times says Flash comes close to these models.

MAI Code 2 is available on Copilot Pro, Pro+, Max, Business and Enterprise plans. Flash is also on the Free and Student plans. Both are available to eligible subscribers on annual plans, GitHub said. The models are billed by usage, and users select them from the model picker, for example in VS Code. On Business and Enterprise plans, the new models are enabled automatically unless an administrator blocks them in Copilot’s model policy.

Previous coverage

Report an error in this story