OpenAI announced that its most advanced artificial intelligence model, the GPT-5.6 series, will be released to the public on Thursday. The company confirmed through a post from San Francisco that the three versions—Soul, Terra, and Luna—will be made globally available, expanding preview access worldwide. Reports indicated that the U.S. government had allowed the large-scale release following technical reviews and meetings between officials and the company.

The GPT-5.6 series follows a limited preview shared with trusted U.S. partners at Washington’s request due to national security concerns. Researchers have expressed unease that the new generation of large language models, including Anthropic’s Mythos series, could identify software vulnerabilities that hackers might exploit. Within the White House, officials remain divided over how much oversight should apply to such powerful systems.

OpenAI stated it is working with Washington to create a repeatable framework for future model releases aligned with cybersecurity directives. The company also said the Terra model will cost half as much as GPT-5.5, a move aimed at retaining customers amid growing competition with Anthropic and Google.