Overview

• On July 12, Sam Altman announced the indefinite postponement of OpenAI’s first open-source reasoning model to allow for further safety testing and high-risk reviews.
• This follows a June postponement and underscores the firm’s shift toward open-weight releases after six years without an open-source model.
• The forthcoming release will let external developers examine and modify the model’s source code and primary functions.
• Once distributed, the model’s weights cannot be universally retracted or patched, heightening the importance of pre-launch governance.
• Chinese competitors such as DeepSeek and Alibaba propelled a global open-source AI movement earlier this year that reshaped market dynamics.