AI News·18 Nov 2026·via It Security News

    Microsoft launches MAI Code 2 models to challenge OpenAI and Anthropic in AI coding

    Microsoft launched two new AI coding models, MAI Code 2 and MAI Code 2 Flash, at Ignite 2026, aiming to challenge OpenAI and Anthropic's offerings. The company stated these models are designed for common enterprise coding tasks, marking an effort to build proprietary AI capabilities and reduce reliance on external partners.

    Microsoft introduced two new AI coding models, MAI Code 2 and MAI Code 2 Flash, at its Ignite 2026 event. MAI Code 2 is engineered for enterprise coding tasks such as deep debugging, implementing multi-file changes, repository-wide reasoning, and user interface development. Its counterpart, MAI Code 2 Flash, focuses on high-volume tasks like code explanations, targeted edits, routine fixes, and rapid iterations. Microsoft stated these models prioritize daily coding work for its customers and are not yet tuned for more complex problems such as fundamental science research.

    This launch represents a strategic move by Microsoft to develop its own AI model capabilities, aiming to lessen its dependence on external providers like OpenAI and Anthropic. Pareekh Jain, CEO at EIIRTrend & Pareekh Consulting, noted that Microsoft gains an advantage through its existing developer platforms, GitHub, Visual Studio, and Azure, which are already widely used. Jain suggested Microsoft's immediate goals include strengthening its developer ecosystem, reducing AI costs, and protecting its profits by potentially locking developers into its own tools.

    Microsoft compared MAI Code 2 against OpenAI's GPT-6.1 Sol and Anthropic's Claude Opus 5.5 across four benchmarks used in GitHub Copilot. MAI Code 2 achieved an average pass rate of 63.2%, slightly below GPT-6.1 Sol's 64.1% and Claude Opus 5.5's 65.3%. On individual benchmarks, MAI Code 2 scored 73.2% on DeepSWE 1.1, 84.6% on Terminal-Bench 2.1, 43.1% on SWE-Atlas, and 52% on FrontierCode v1.1. For MAI Code 2 Flash, comparisons with GPT-6 Luna and Claude Haiku 5.5 across three benchmarks showed an average pass rate of 61%, outperforming GPT-6 Luna at 55.1% and Claude Haiku 5.5 at 53.9%.

    Beyond performance, Microsoft highlighted cost efficiency. The company claimed an average cost per task of $0.309 for MAI Code 2, significantly lower than GPT-6.1 Sol's $0.761 and Claude Opus 5.5's $6.306. Similarly, MAI Code 2 Flash was presented with an average cost per task of $0.052, compared to GPT-6 Luna's $0.080 and Claude Haiku 5.5's $0.319. However, these reported costs for MAI Code 2 and Flash are based on a 50% promotional pricing structure.

    The promotional pricing for MAI Code 2 is $0.50 per million input tokens, $0.025 per million cached input tokens, and $2.50 per million output tokens. This promotional rate is available until May 17, 2027, after which prices will increase to $1.00, $0.05, and $5.00 per million tokens, respectively. MAI Code 2 Flash also has promotional pricing of $0.05 per million input tokens, $0.005 per million cached input tokens, and $0.25 per million output tokens, which will increase to $0.10, $0.01, and $0.50 per million tokens after the promotional period.

    Share this article

    Go deeper
    Original reporting by It Security NewsWe don't republish, read the full story →

    Related reading

    6 stories