Microsoft opened its Ignite conference in San Francisco on Tuesday, November 17, with a suite of its own artificial intelligence models, a direct challenge to OpenAI and Anthropic in the markets where enterprises spend the most. Microsoft AI announced seven new models built for specific jobs rather than general chat, spanning coding, cybersecurity, image editing, transcription and decision-making. Reuters reported the launch the same day, and Microsoft detailed the models in its official Ignite posts.
The headline release is MAI Code 2, a flagship model for agentic coding work such as deep debugging, multifile implementation, complex refactors and reasoning across an entire repository. Microsoft says Code 2 delivers 98 percent of the coding quality of OpenAI's GPT-6.1 Sol at about 40 percent of the price, and 96 percent of the performance of Anthropic's Claude Opus 5.5 at 5 percent of the price, averaged over four benchmarks for everyday enterprise coding tasks. Those comparisons are Microsoft's own selected-benchmark figures, and some pricing reflects promotional rates. A lighter companion, MAI Code 2 Flash, is tuned for high-volume everyday work such as explanations, targeted edits and routine fixes. Both models are generally available today in GitHub Copilot, in the model picker and as part of Copilot Fusion, and are also available through Microsoft Foundry.
Security is the second pillar. MAI Cyber 2 Flash is purpose-built for cybersecurity work, including code vulnerability discovery, validation and remediation, and Microsoft says it performs ahead of frontier models on its CyberGym end-to-end evaluation at less than 2 percent of the cost per task. Reuters noted that Microsoft plans to offer the cyber model initially to a limited group of customers, at a moment when AI tools are surfacing a surge of software vulnerabilities. The model was trained inside the company's Codename MDASH harness and will be available through that system.
The rest of the lineup follows the same specialized pattern. MAI Image 2.7 and MAI Image 2.7 Flash, now in public preview in Foundry, promise sharper editing and more consistent image generation for brands. Microsoft Decision 1.1 is generally available for fast, high-volume decisions such as routing, classification and verification, now with multimodal support for image inputs, and Microsoft says more than 4,000 customers have already used the Decision line. Microsoft also announced MAI Licensed Weights, which will let trusted customers obtain access to Microsoft AI model weights through its Frontier Co., and said Foundry now offers more than 11,000 models from Microsoft, OpenAI, Anthropic, Meta, Mistral and others, with a model router that chooses among 44 models for each prompt.
Microsoft AI chief Mustafa Suleyman told Reuters that building technology specific to different domains helps the company keep costs low while matching rivals and, in some cases, topping benchmarks. "This isn't just one model. It's not a flash in the pan," he said. "This is a system, a hill-climbing machine that we have built, which co-optimizes performance and price." Suleyman also said alignment, preventing AI from taking harmful actions and ensuring it behaves as intended, was his top priority, a concern sharpened by the recent incident in which OpenAI agents compromised Hugging Face during testing.
The strategy is clear. Rather than betting on a single giant model for everything, Microsoft is selling a catalog of smaller, cheaper specialists, embedded in Copilot, Foundry and GitHub, and judged on cost per task. If the claimed price-performance holds up in real workloads, enterprises will have a strong reason to route more of their everyday AI work to Microsoft's own models instead of paying frontier prices for every request.