Open AI Models Gain Enterprise Ground, Cutting Catch-Up Time by Half
Updated
Updated · Computerworld · Aug 26
Open AI Models Gain Enterprise Ground, Cutting Catch-Up Time by Half
3 articles · Updated · Computerworld · Aug 26
Summary
Open models are winning more enterprise deployments for reasoning, agentic and physical AI because companies can fine-tune them to internal workflows, keep data on-premises and avoid paying for broad frontier models they do not need.
SemiAnalysis said each new generation of open-source models now takes half as long to catch up with the first closed-source model of its era, underscoring how quickly the performance gap is narrowing.
Open-weight models have become the most common enterprise option, giving companies visibility into model parameters and deployment control, though they still withhold elements such as code and training data that fully open-source models disclose.
China's DeepSeek, Alibaba's Qwen and Moonshot's Kimi have helped accelerate adoption, while countries including Germany, France and India are also backing open models for sovereignty and local regulatory needs.
The shift still carries trade-offs: enterprises often must handle maintenance and updates themselves, and poorly vetted open models or agents can create new security and data-leak risks.