AI large-scale models, exemplified by GPT and DeepSeek, have had a profound impact on the global order. Extending the “system-lifeworld” theory to the large language models not only uncovers how the two technological trajectories of closed-source hegemony and open-source cooperation differently construct the global order, but also achieves a paradigmatic shift of the theory from a critique of modernity to an interpretation of the global digital order. Closed-source large models epitomized by GPT are progressively reinforcing U.S.-centric global hegemony, seeking to establish a worldwide “digital System”. Conversely, open-source counterparts like DeepSeek emerge as transformative technological forces facilitating globally negotiated construction, thereby endeavoring to shape a global “digital lifeworld”. The GPT technology endeavors to construct a global “digital system” that consolidates technological and cultural hegemony, while eroding the communicative rationality among pluralistic actors within intersubjective interactions. Its technical training paradigm imposes exceptionally high demands for data accumulation and computational reserves, forging a monopolistic developmental trajectory for large-scale models. Concurrently, AI oligopolies leverage GPT to advance “uni-acculturation” of knowledge across global domains, thereby perpetuating the reproduction of knowledge imperialism. DeepSeek, however, functions as a transformative force. DeepSeek facilitates the global development of large-scale models towards “co-ownership, co-governance, and sharing”, promoting “trans-acculturation” of knowledge and negotiated construction within the global digital space. This fosters communicative rationality among pluralistic actors across technological and cultural “dual fields”, thereby constructing a global “digital lifeworld”.