10th July 2026 – (San Francisco) OpenAI has announced the public release of GPT‑5.6, its newest generation of artificial intelligence models, following a delay last month requested by the United States government over cybersecurity considerations.
The San Francisco‑based company confirmed on Thursday that the GPT‑5.6 range comprises three distinct models. “Sol” is positioned as the flagship system, “Terra” is designed as a balanced option for routine professional tasks, and “Luna” is described as the fastest and most cost‑effective variant.
According to OpenAI, GPT‑5.6 Sol delivers advances in coding, research, cybersecurity and scientific applications, while operating with improved efficiency and lower token usage than both its predecessors and rival frontier models.
The company has also introduced two enhanced performance modes. The “max” configuration allows the system additional time to analyse and refine its reasoning, whereas the “ultra” mode deploys multiple agents simultaneously to tackle more complex assignments.
GPT‑5.6 became accessible on Thursday through ChatGPT, Codex and the OpenAI API, with the worldwide rollout expected to continue over the following 24 hours.
OpenAI stated that the new models are being released with what it described as its strongest safety measures to date. While GPT‑5.6 demonstrates greater capability in areas such as biology and cybersecurity compared with earlier versions, the firm said it remains below the “critical” risk threshold defined in its preparedness framework. The models were subjected to what the company characterised as its most comprehensive safety testing prior to launch.