Microsoft is reportedly working on adding non-OpenAI artificial intelligence (AI) models to its 365 Copilot products. According to the report, the Redmond-based tech giant is trying to reduce its reliance on OpenAI’s AI models due to concerns related to speed and cost. To diversify, the company is said to be considering third-party models as well as internally developed models. Notably, Microsoft introduced Copilot 365 in March 2023 and one of its unique selling points (USP) was integration with GPT-4 AI models.
Microsoft aims to reduce reliance on OpenAI AI models
According to a Reuters report, Microsoft is working on bringing internal and third-party AI models to its 365 Copilot products. Citing sources familiar with the matter, the publication claimed that the Windows maker is now actively trying not to rely solely on OpenAI’s large language model (LLM) to offer AI solutions to its enterprise customers.
If true, it would be a major departure from Microsoft’s current AI strategy. The company invested $1 billion (roughly Rs. 8,519 crore) in OpenAI in 2019, followed by $10 billion (roughly Rs. 85,195 crore) in 2023 as part of an ongoing partnership with the AI firm. The partnership allows Microsoft early access to all AI models developed by the firm.
According to the report, the tech giant’s biggest concern at the time is the cost and speed of OpenAI-developed AI models. This is a significant sticking point because Microsoft does not use these AI models for internal use, but the technology is used to build AI products for its enterprise customers.
“We incorporate different models from OpenAI and Microsoft depending on the product and experience,” Microsoft told Reuters.
To diversify, the tech giant is reportedly looking at ways to develop third-party AI models as well as in-house LLMs. The company has released several small language models (SLMs), the latest of which is the open-source Phi-4 AI model. However, it has yet to develop a general-purpose LLM.
Recently, Microsoft began expanding its GitHub Copilot by allowing developers to access Google’s Gemini and Anthropic’s cloud AI models. Notably, GitHub Copilot was the first AI product released by the company after its collaboration with OpenAI. It’s possible that other 365 CoPilot devices may also allow users to switch between different models in the future.
