Skip to content
Brief The AI-First Web ·

Microsoft unveils six AI models; says its coding model is far cheaper

Microsoft says its MAI Code 2 coding model nears rival performance at lower prices, on benchmarks it chose.

In brief
  • Microsoft says its coding model MAI Code 2 reaches 98% of GPT-6.1 Sol's performance at 40% of the price. OpenAI announced GPT-6.2 Astra and GPT-6.2 Sol the same day.
  • Microsoft chose the benchmarks, and some prices are promotional discounts. Customised model weights must run inside Azure.

At its Ignite conference in San Francisco on Tuesday, Microsoft announced six AI models for coding, cybersecurity, image making and voice, Axios reported. Microsoft says its coding model, MAI Code 2, reaches 98% of GPT-6.1 Sol's performance while costing 40% as much. Against Claude Opus 5.5, it claims 96% of the performance for 5% of the cost. OpenAI also announced GPT-6.2 Astra and GPT-6.2 Sol that day.

Microsoft AI chief Mustafa Suleyman told Axios that specialised models hold more knowledge in the same number of parameters, the values a model learns in training. Axios noted the price of that focus: a model tuned for everyday coding will be weaker at advanced maths and science. Microsoft picked the benchmarks, and some of its price gaps reflect promotional discounts. It gave no release date for a general-purpose model and no target for Copilot traffic.

A narrow model spends its limited capacity on one job, so it can match a larger model there and lose ground elsewhere. That moves the buying decision from picking one model to routing each kind of work to the model that suits it. Ask which of your workloads are only coding, what the price is without the promotion, and what running licensed weights only inside Azure means for your data.

A WebPulse Brief: a short report of an important event, written by the WebPulse Newsroom with AI assistance and checked against the reporting below. When there is more to explain, we follow up with a full story. How we use AI.

Reporting: Axios.