OpenAI releases GPT-5.6 publicly after US government safety review
OpenAI is making its GPT-5.6 model family — the flagship Sol, the mid-tier Terra, and the lightweight Luna — broadly available after the US Commerce Department cleared the models for wide release on July 8. The company had previewed the family on June 26 but, at the government’s request, initially shipped it to roughly 20 trusted partner organizations while a safety review proceeded.
The gating point was cybersecurity. OpenAI has described GPT-5.6 Sol as its most capable model yet for security-related tasks, including vulnerability research and exploitation — the same capabilities that make it useful for defenders and attractive to attackers. The two-week hold reflects a now-familiar pattern in which frontier models undergo government review over dual-use risk before general release.
The clearance mirrors recent handling of other US frontier labs whose top models were briefly restricted on similar grounds. For customers, the practical change is access: developers and enterprises can now build on the full GPT-5.6 line rather than wait behind the partner-only preview.