Microsoft cuts reliance on ChatGPT, Claude over AI costs
What's the story
Microsoft is reportedly shifting its focus toward its own AI models, in a bid to cut down on the rising costs of using advanced AI at a corporate level. According to Bloomberg, the tech giant has started using its own MAI models for tens of thousands of AI prompts every week in Excel and Outlook software. This shift comes as a response to the increasing expense of AI tokens, which measure how much computing work a model is doing.
Model transition
Copilot still relies on external models
The transition to MAI models has seen Microsoft software, like Excel and Outlook, rely less on the models from OpenAI and Anthropic. However, Bloomberg points out that this is just a small part of Microsoft's overall AI usage. The company's workplace AI assistant Copilot still requires a huge amount of AI tokens for its operations.
Model launch
Microsoft's in-house models
The news of Microsoft's shift comes a little over a month after it announced seven new in-house models, including MAI-Thinking-1. The company had said that this model was designed for "high efficiency and performance, but importantly, at a low-token cost." Microsoft describes MAI-Thinking-1 as a mid-sized model with 35 billion active parameters and a 256K context window. In blind tests, it matched the coding abilities of Anthropic's popular Claude Opus 4.6.
Cost comparison
Industry trend toward cost-effective AI solutions
The launch of Microsoft's new models comes as the industry is increasingly turning its attention toward cheaper and more efficient AI models. China's DeepSeek made headlines earlier this year with budget-friendly models like V4-Pro, which costs $0.435 per million input tokens and $0.87 per million output tokens. In comparison, Anthropic's most advanced model Fable 5 costs $10 per million input tokens and $50 per million output tokens.
Cost concerns
Microsoft's concerns over high AI costs
Microsoft AI CEO Mustafa Suleyman has already expressed concerns over the high costs of using advanced AI models. He told Bloomberg last month that "Anthropic is extremely expensive and I think many people are urgently looking for alternatives." He added, "We pay a lot of money to Anthropic, so our goal is to reduce and ultimately eliminate that cost."