Microsoft is reportedly seeking to diversify the AI models used in its Microsoft 365 Copilot, moving away from an exclusive reliance on OpenAI‘s technology. According to a report by Reuters, the tech giant is looking into both in-house and third-party AI models to improve efficiency and reduce costs associated with running its AI-powered assistant.

The shift comes as Microsoft grapples with the high costs and speed limitations of relying solely on OpenAI’s models. Sources close to the matter told Reuters that Microsoft is not only developing its own smaller language models (SLMs), like the recently released open-source Phi-4, but is also customizing other open-weight models. This strategy aims to make Microsoft 365 Copilot more cost-effective and potentially lower the price for users.

Historically, Microsoft has had a deep partnership with OpenAI, having invested billions to secure an exclusive license to use OpenAI’s GPT-4 model within Microsoft 365 Copilot. This AI assistant, which comes at an additional $30 per user per month on top of a Microsoft 365 subscription, integrates into various Microsoft applications, enhancing user productivity by automating tasks and providing real-time assistance.

However, the landscape is changing. A statement from a Microsoft spokesperson to Reuters confirmed, “We incorporate various models from OpenAI and Microsoft depending on the product and experience,” suggesting a broadening of its AI toolkit. Despite this, they emphasized that “OpenAI remains Microsoft’s main partner to work on the most advanced AI models coming to the market.”

This strategic pivot was hinted at with the establishment of “Microsoft AI,” a new division led by Mustafa Suleyman, co-founder of DeepMind and Inflection. This division is focused on developing proprietary AI technologies, which have already been incorporated into the consumer version of Microsoft Copilot launched in October. This version uses a blend of Microsoft’s own AI models alongside those from OpenAI, indicating a significant shift in Microsoft’s AI deployment strategy.

Furthermore, the involvement of Microsoft’s top brass, including CEO Satya Nadella, in overseeing these changes underscores the company’s commitment to this new direction. The integration of alternative AI models isn’t limited to Microsoft 365 Copilot; GitHub Copilot, another Microsoft AI tool, has also broadened its AI model support to include offerings from Anthropic and Google, as announced in October. This reflects Microsoft’s acknowledgment of the need for flexibility and choice in AI model usage, allowing developers to select models that best suit their needs.

The implications of these developments are multifaceted. For Microsoft, reducing dependency on external AI technology could lead to more control over its AI-driven products, potentially enhancing performance and reducing costs. For users, this could translate into more tailored AI experiences and possibly more competitive pricing. However, this shift also prompts questions about the future of Microsoft’s relationship with OpenAI and how this diversification might affect the capabilities and quality of AI assistance provided.