OpenAI has announced the limited preview release of the GPT-5.6 series, comprising three variants: the flagship Sol, the balanced Terra, and the fast, affordable Luna. The rollout has been deliberately restricted at the request of the Trump administration, which asked to limit access for security reasons. The three models are currently available only to a small group of trusted partners and organizations via API and Codex, with a broader release expected in the coming weeks.
Sol is the most powerful model OpenAI has ever built
GPT-5.6 Sol boasts significant improvements in agentic capabilities, particularly in coding, biology, and cybersecurity. OpenAI has introduced a new reasoning effort called "max" and an "ultra" mode that uses sub-agents for complex tasks. According to the company, Sol is better at helping developers find and fix vulnerabilities than carrying out end-to-end attacks, marking a major step forward in cyber defense.
Sponsored Protocol
Terra delivers GPT-5.5 performance at half the price
The Terra model is designed for everyday use: its performance is comparable to GPT-5.5, but at half the cost. Luna, on the other hand, is the most affordable model in the lineup, offering what OpenAI describes as "strong capability" at its lowest price point. This tiered pricing strategy aims to serve both enterprise needs and independent developers.
White House imposes limits following June executive order
The Trump administration required OpenAI to hold back general release of GPT-5.6, following a new benchmarking and evaluation process for AI models established by a June 2 executive order. OpenAI, while accepting the temporary limitation, expressed concern that this government access process should not become the norm. In an official statement, the company said: "We do not believe this kind of government access process should become the long-term default. It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them." The tension between innovation and regulation fits into a broader context: OpenAI had already released GPT 5.6 in staged manner at White House request over security fears, and Europe is accelerating its AI sovereignty push to reduce dependence on US models. US restrictions and talent exodus are driving the EU to build its own models.
Sponsored Protocol
Enhanced safety with testing against real-world attacks
OpenAI claims that GPT-5.6 Sol has its "most robust safety stack to date," with safeguards for high-risk activity, sensitive cyber requests, and misuse. The model has been tested for vulnerabilities and hardened against real-world attacks. The safeguards allow delivering substantial benefit for legitimate defensive work while limiting prohibited offensive use. For more details on the safety framework, visit the Wikipedia page on OpenAI.
Sponsored Protocol
Source: https://www.macrumors.com/2026/06/26/openai-gpt-5-6-sol
