Altman: GPT-5.6 is 54% more token efficient on agentic coding (www.cnbc.com)

OpenAI has launched its latest AI model, GPT-5.6 Sol, along with its companions Terra and Luna, after initially rolling it out to a limited group of trusted partners at the behest of the U.S. government. CEO Sam Altman announced that GPT-5.6 Sol is 54% more token efficient for agentic coding tasks than its predecessors, positioning it as a strong contender against competing models. This increased efficiency is crucial for enterprises looking to optimize their AI expenditure and maximize value in coding applications. The rollout involved collaboration between OpenAI and top U.S. government officials, ensuring regulatory compliance and safety assurances. This cooperative approach highlights the government's keen interest in overseeing powerful AI advancements, which Altman notes is vital for maintaining public confidence in AI technologies. By prioritizing safety and efficiency, OpenAI aims to provide broader access to its powerful models, addressing potential concerns related to their deployment in real-world applications. The significance of GPT-5.6 Sol lies not only in its technical improvements but also in its strategic alignment with regulatory frameworks that govern AI technology.