Why Microsoft Thinks Smaller, Cheaper AI Can Beat Claude and GPT
Microsoft introduced specialized AI models for coding, cybersecurity and image generation, positioning them as lower-cost alternatives to offerings from Anthropic and OpenAI. It claims MAI Code 2 approaches leading models on enterprise coding tests at a fraction of their cost, while MAI Cyber 2 Flash performed slightly better than GPT-6 Astra on a vulnerability-discovery benchmark for under 2% of the cost. The claims rely partly on Microsoft-run evaluations and discounted pricing, and some comparisons are absent or against older models. MAI Image 2.7 leads a third-party image-editing preference ranking by a narrow margin, with substantially lower quoted prices than OpenAI’s top-quality option. Code 2 is available in GitHub Copilot; Cyber 2 Flash is due to roll out to eligible customers in mid-December. Microsoft also offers selected businesses licensed model weights for deployment in Azure.
