Microsoft is reportedly evaluating DeepSeek’s open-source V4 model for Copilot Cowork as enterprises seek lower-cost AI options, signalling growing acceptance of open-source models in production AI deployments.
Microsoft is reportedly considering integrating DeepSeek’s open-source V4 model into its enterprise AI agent platform, Copilot Cowork, marking a significant endorsement of open-source AI in enterprise environments. According to Axios, the Chinese model is being evaluated as a lower-cost alternative alongside offerings from Anthropic and OpenAI as Microsoft shifts Copilot Cowork from a flat-rate to a pay-per-use pricing model.
The move reflects a broader industry trend towards multi-model AI strategies driven by rising AI agent costs. As AI agents increasingly perform autonomous reasoning and task execution, token consumption has surged, prompting enterprises to balance performance with affordability. A tech industry insider noted, “Jevons’ paradox, where lower prices lead to explosive usage, is replaying in the AI market.”
Cost is a major factor behind DeepSeek’s growing appeal. Artificial Analysis estimates DeepSeek V4 Pro costs just $0.05 per task and V4 Flash $0.03, compared with $1.07 for GPT-5.5 and $2.05 for Claude Opus 4.8. While DeepSeek trails leading proprietary models in benchmark scores, industry observers view the performance trade-off as acceptable given the substantial savings.
Citigroup described the trend as “not an overstatement to label it ‘DeepSeek Shock 2,’” adding that “AI models are rapidly becoming commoditized, and companies will build multi-model AI systems.”
The rise of DeepSeek and other Chinese open-source models also reflects improving capabilities, growing real-world adoption, and increasing demand for sovereign AI, underscoring how open-source AI is moving from experimentation to mainstream enterprise deployment.
