A New Model Family Arrives on Thursday

OpenAI is rolling out its next generation of AI models this week. CEO Sam Altman confirmed in a Wednesday post on X that the GPT-5.6 lineup, made up of three distinct models, launches on Thursday. The family had already been previewed back on June 26, but OpenAI chose to delay the wider release rather than push it out immediately after the announcement.

Exact availability details for each model are still unclear. What's confirmed is that all three versions are set to go live on the same day.

Meet the Three Models: Sol, Terra, and Luna

GPT-5.6 Sol: The Flagship

Sol sits at the top of the GPT-5.6 family. It's built to take on demanding reasoning work and to oversee AI agents. According to OpenAI, Sol also cuts down on factual errors, or hallucinations, compared to the company's earlier models — a meaningful upgrade for anyone relying on ChatGPT for accuracy-sensitive tasks.

GPT-5.6 Terra: The Balanced Option

Terra lands in the middle of the lineup, striking a balance between how capable the model is and what it costs to run. It's positioned as the practical middle ground for users who don't need Sol's full reasoning power but want more than the entry-level option.

GPT-5.6 Luna: Built for Speed

Luna is the fastest and least expensive model of the three. It's the go-to choice for tasks where speed and cost matter more than deep reasoning.

Stronger Cybersecurity, Tighter Safety Guardrails

OpenAI has described the GPT-5.6 family as representing a meaningful step up in cybersecurity capability. That added power comes paired with more advanced safety guardrails, according to details the company laid out when it first previewed the model family in June.

Why OpenAI Slow-Rolled the Release

Even though the GPT-5.6 family was announced publicly back on June 26, OpenAI deliberately held off on releasing it right away. The company isn't alone in taking this cautious approach — Anthropic has followed a similar path with Claude Mythos 5, another model described as being among the most advanced AI systems available. As models reach this level of capability, the companies behind them appear to be treating public release with more hesitation than in the past.

Washington Steps Into Frontier AI Oversight

A New Executive Order

For the first time, the U.S. government is directly involved in reviewing frontier AI models before they reach the public. On June 2, President Donald Trump signed an executive order directing the Defense Department to build a system through which AI developers voluntarily give the government access to new frontier models. Once that access is granted, government officials get a 30-day window to review the technology and flag any concerns.

This is a separate legal framework from the one that led to the government temporarily shutting down and later restoring access to Anthropic's Claude Fable 5, which involved export control rules rather than this new voluntary review process.

Early Access for Unnamed "Trusted Partners"

OpenAI has said it voluntarily gave both the government and a group of "trusted partners" early access to GPT-5.6 ahead of Thursday's release. The company hasn't disclosed who those partners are or what expertise they bring to the review, and it declined to go into specifics beyond pointing back to its original announcement.

The White House Weighs In

A White House official clarified that the government did not formally approve OpenAI's release of GPT-5.6, and also noted that such approval isn't a requirement for public release under the new executive order. Officials stressed that participation in the review process is entirely voluntary on OpenAI's part.

A Shift in the Government's Approach to AI

This level of government involvement marks a notable change in direction. The Trump administration had previously argued strongly against AI regulation, warning that added oversight could slow innovation and risk the U.S. falling behind China in the global AI race. Now, with this new review process in place, the pace at which frontier models reach the public — and how the U.S. is perceived on the world stage technologically — is shifting in real time.