OpenAI was not directly accessible to all users when the GPT-5.6 series was launched. According to the company, at the request of the United States Government, the new model provides a preview to a small number of partners, with subsequent and gradual expansion. This indicates that the management of the front-line AI model in the United States has been extended to the release chain.

Three models on line.

The release included three versions, Sol, Terra and Luna, which corresponded to high performance, balanced cost and high-swallowing scenes, respectively. OpenAI states that the new model has been upgraded in terms of software engineering, Agent missions and cybersecurity defence research.

  • Sol for complex reasoning, research and development missions
  • Terra, play business and daily productivity.
  • Luna Focus on speed and cost for large-scale services

In it, the flagship Sol added a higher-level mode for longer-term reasoning and multi-Agent synergy. OpenAI states that the model completed the automated security test and external red team assessment prior to its release.

The government wants a limited preview.

The market is more concerned with this change in the way it goes online than with its improved performance. In its bulletin, OpenAI disclosed that the company had demonstrated model capabilities to the United States Government prior to its official release and had opened previews to a small number of “credible partners” as required.

According to the report, some 20 of the first institutions to be granted competence will gradually expand in the coming weeks. The United States Government hopes to establish a unified security assessment process, focusing on the potential network security risks of the model, before the front-line model becomes fully available.

OpenAI opposes the long-term approval system

While complementing this limited opening, OpenAI made it clear that government intervention in model visits should not be a long-term default. According to the company, the broad access to advanced models remains a goal and the current arrangements are short-term measures.

OpenAI believes that the speed of access to tools by developers, businesses, scientific institutions and cyber-security defenders is affected by the long-term application of case-by-case approvals. The company would like to take advantage of this preview to establish a repertoireable distribution process with the Government.

ChatGPT will gradually switch

OpenAI also indicated that in addition to API, ChatGPT and Codex follow-up will gradually access GPT-5.6 and the developer platform will later open more access. At the same time, the company states that the GPT-5.6, although enhanced in terms of autonomous missions and gap analysis, has not yet reached the risk threshold requiring the highest level of qualification.