In a recent article, Wired reports that OpenAI’s first open-weight artificial-intelligence models in years are being evaluated for use by the U.S. Department of Defense and its contractors. The models, named gpt-oss-120b and gpt-oss-20b, mark a significant shift from OpenAI’s previous strategy of keeping model weights proprietary. Because they can be run locally—without requiring an internet connection—they open up possibilities for high-security use cases such as military systems or government installations.
However, the article also highlights that while these models are now available for local deployment, some defense-industry insiders believe they still lag behind certain competitors in performance and capability. OpenAI acknowledges that open-weight releases entail different trade-offs compared to purely proprietary models—specifically in terms of control, security and fine-tuning—and show that the company is navigating a new balance between openness and oversight.
From a strategic perspective, this move by OpenAI has several implications: it signals to both governments and enterprises that advanced AI need not remain locked behind cloud APIs; it opens up new deployment options in environments with strict data-sovereignty or offline requirements; and it intensifies competition in the open-weight model space, especially with rivals such as Meta Platforms and Chinese firms.
