Microsoft, now free of the shackles that previously tied it to OpenAI, apparently has eyes for other AI models—particularly ones that are cheap. According to a report from Axios, tokenmaxxing is taking its toll on Microsoft’s AI tool Copilot, and it is looking at the possibility of using its own version of the Chinese open-source model DeepSeek to keep its costs down. Assuming Microsoft were to take this path, the company would reportedly use a modified and self-hosted version of DeepSeek-V4, the latest version of the popular model from China. It would be used as a lower-cost option to power the company’s Copilot Cowork, the agentic AI offered through the Microsoft 365 enterprise suite. It’s certainly not hard to understand why Microsoft would want to cut some costs. The company’s Copilot Cowork currently runs on Anthropic’s models and has versions that are compatible with OpenAI’s options. But both of those keep raising their prices and have moved away from “all-you-can-eat” pricing that was starting to cost them. That has pushed Microsoft to move to a usage-based pricing model for its agentic AI solution, per Axios. If it can keep the costs down with a cheaper model, it’ll keep its customers happy.