Ecosystem

MAI Code 2: Microsoft claims 96 percent of Claude Opus 5.5 at 5 percent of the price

2 min read AI-generated

Five more models come with it. Companies can license the weights and tune them on their own data, as long as they run them in Azure.

Featured image for "MAI Code 2: Microsoft claims 96 percent of Claude Opus 5.5 at 5 percent of the price"

Microsoft announced six of its own AI models on November 17 at its Ignite conference in San Francisco. They cover coding, cybersecurity, images and voice. Microsoft AI CEO Mustafa Suleyman told Axios they can compete with OpenAI and Anthropic at a fraction of the cost.

The models

MAI Code 2 is the coding model. Microsoft says it delivers 98 percent of GPT-6.1 Sol’s performance at 40 percent of the price, and 96 percent of Claude Opus 5.5’s at 5 percent of the cost.

MAI Code 2 Flash is the smaller, faster version. It’s supposed to match GPT-6 Luna at 65 percent of the price and Claude Haiku 5.5 at a fifth of the cost.

MAI Cyber 2 Flash beats GPT-6 Astra on a security benchmark at under 2 percent of the cost per task, according to Microsoft. There’s also MAI Image 2.7, which Microsoft says ranks first in image editing, plus new models for transcription and real-time voice.

Axios points out that the comparisons rely on selected benchmarks, and some of the price advantage comes from promotional discounts.

Built for one job

Suleyman’s recipe: a model tuned for one task can pack more knowledge for that task into the same number of parameters. The trade-off is that a model built for everyday coding is weaker at advanced math or research. Alongside this, Microsoft is working on its own general-purpose frontier model. Suleyman says pretraining results are state of the art and reinforcement learning is coming along well, but he gave no release date. “My team is building true superintelligence,” he said.

Companies can license the models’ weights and customize them with their own data, but they have to deploy them in Azure. Suleyman says it should “feel like open source.”

In Copilot, Microsoft sets the default

MAI Code 2 is coming to GitHub Copilot. Claude models are among the options there today. Suleyman wouldn’t say how much Copilot traffic he expects to move to Microsoft’s own models, only that he predicts significant adoption. If Microsoft offers a model inside its own tool that, by its own math, does 96 percent of what Opus 5.5 does for a twentieth of the price, Anthropic is competing against the host. If you use Copilot, it’s worth checking which model is actually selected from now on.

Sources

MicrosoftMAI Code 2GitHub CopilotClaude Opus 5.5Mustafa Suleyman