Microsoft is reportedly evaluating Kimi AI’s Kimi K3 open-weight large language model, built by China’s Moonshot AI, for potential integration into Copilot, while also preparing to make the model available through Azure AI. The evaluation focuses on AI inference efficiency, coding and reasoning performance, cost per token, compatibility with Copilot services, and enterprise deployment through Azure AI. Microsoft is testing whether Kimi K3 can handle some Copilot tasks currently powered by OpenAI models, reflecting a broader strategy of expanding beyond exclusive reliance on OpenAI models to offer customers a wider portfolio and potentially reduce inference costs. No final deployment decision has been announced. The evaluation comes as ChatGPT’s lead has narrowed to about 58% as rivals gain share. Microsoft has steadily expanded Azure AI beyond OpenAI models by adding foundation models from multiple developers, positioning Azure as a model marketplace where customers can select models based on cost, latency, performance, and regulatory considerations.