The delay comes amid fierce competition among AI developers to boost model performance, cut costs and expand enterprise capabilities, fueling a steady, industrywide stream of new systems and reasoning models.