Copilot may soon receive more AI models from Microsoft and reduce the presence of ChatGPT.
It has been reported that Microsoft is considering integrating more of its own artificial intelligence models into Copilot, which would allow it to reduce its dependence on OpenAI. Additionally, it is exploring competitive options such as DeepSeek and Meta.
Microsoft has been guiding many of its artificial intelligence efforts through its partnership with OpenAI, highlighting products like Copilot that use the most advanced ChatGPT models. However, the company appears determined to develop its own AI models within its popular software suite, in addition to advancing the creation of an alternative to OpenAI's reasoning models, part of the "GPT-o" family.
Recently, the team from Microsoft's AI unit completed the training of a new family of artificial intelligence models, currently being developed under the codename “MAI.” Employees within this unit hope that these internal models can compete in performance with industry leaders like OpenAI and Anthropic. Under the direction of Mustafa Suleyman, Microsoft is embarking on this initiative to reduce its dependence on OpenAI and build its own artificial intelligence ecosystem for Copilot applications.
In recent times, Microsoft has been consolidating its own AI infrastructure. At the end of February, the company introduced new smaller language models called Phi-4-multimodal and Phi-4-mini, which feature multimodal capabilities, allowing for text, voice, and vision processing, similar to what OpenAI's ChatGPT and Google's Gemini offer. These models are already available to developers through Azure AI Foundry and third-party platforms like HuggingFace and NVIDIA's API Catalog. In tests conducted, the Phi-4 model has outperformed Google’s latest Gemini 2.0 models across several evaluation metrics.
Moreover, Microsoft is planning to commercially launch its “MAI” models through its Azure service. The company is also exploring third-party options, including DeepSeek, xAI, and Meta, which aim to provide high-performance models at reduced development costs. Recently, DeepSeek was noted for achieving very favorable cost-benefit relationships.
On the other hand, Microsoft is not only focused on replacing OpenAI's GPT infrastructure for Copilot but is also working on its own reasoning models, which could put it in direct competition with products like OpenAI's GPT-o1 and startups such as DeepSeek, which provide reasoning capabilities. This development has taken on greater urgency due to tensions between the teams at Microsoft and OpenAI, stemming from OpenAI’s lack of transparency regarding the operational details of its models, such as GPT-o1.
Reasoning models are considered the next frontier in AI development, as they offer a more nuanced understanding of queries, logical deduction, and enhanced problem-solving capabilities. Microsoft claims that its Phi-4 model delivers an improvement in linguistic, mathematical, and visual scientific reasoning.
