OpenAI launched GPT-5.6 on June 26, 2026, in a limited preview restricted to approximately 20 US government-vetted organizations. General access is expected within weeks, but no firm date has been set.
OpenAI deferred the public rollout of GPT-5.6 on Friday, June 26, 2026, complying with a US government request to limit early access to frontier AI models. The company launched its most powerful model to date a three-tier family called Sol, Terra, and Luna but restricted access to roughly 20 organizations whose participation has been formally shared with federal authorities. It is the first time Washington has directly shaped the release of a frontier AI model before it reached the public.
The decision follows President Donald Trump’s June 2 executive order titled “Promoting Advanced Artificial Intelligence Innovation and Security,” which asks AI developers to voluntarily give the federal government up to 30 days of early access to their most advanced models before releasing them to trusted partners. In practice, that voluntary framework is beginning to look less voluntary. OpenAI cooperated but made clear it was not entirely happy about it.
- What Happened: OpenAI’s GPT-5.6 Launch Explained
OpenAI announced GPT-5.6 on June 26, 2026, introducing a new naming system that replaced single-model releases with three distinct capability tiers. Sol is the flagship, designed for the hardest reasoning tasks including advanced coding, scientific research, and cybersecurity. Terra is a balanced everyday option, offering GPT-5.5-level performance at roughly half the cost. Luna is the fastest and most affordable, built for high-volume routine automation.
The company also introduced Sol Ultra, a compute-intensive high-effort mode that uses multiple sub-agents working in parallel. On Terminal-Bench 2.1, Sol Ultra scored 91.9% — outpacing Anthropic’s Claude Mythos 5 at 84.3% and Claude Opus 4.8 at 78.9%. These are the highest benchmark results ever published by the lab.
- Why the US Government Blocked the Full Public Launch
The restricted rollout stems from a direct request by the White House’s Office of the National Cyber Director (ONCD) and the Office of Science and Technology Policy (OSTP). Both offices asked OpenAI to limit the launch while the Trump administration develops a formal framework for evaluating frontier AI models before public release.
The concern is specific. Advanced AI models particularly those capable of identifying and exploiting software vulnerabilities are now seen as a national security variable. Washington wants the ability to assess those capabilities before bad actors can access them.
Trump’s June 2 Executive Order: The Legal Foundation
The authority behind this request derives from Trump’s executive order “Promoting Advanced Artificial Intelligence Innovation and Security,” signed June 2, 2026. The order asks AI companies to voluntarily give the federal government up to 30 days of access to “covered frontier models” before releasing them to other trusted partners. Federal agencies were given until August 1, 2026, to finalize the framework for what qualifies as a covered frontier model and how the review process will work.
“We are taking this short-term step because we believe it is the strongest path to broader availability in the coming weeks, while we work with the Administration to develop the cyber Executive Order framework and a repeatable process for future model releases.”
— OpenAI, official blog post, June 26, 2026
The order is technically voluntary. It explicitly bars the creation of a mandatory government licensing or pre-clearance requirement. However, critics argue that what is voluntary on paper becomes practically compulsory when a company is seeking government contracts, operating in regulated industries, or simply trying to maintain good relations with Washington.
OpenAI Pushes Back — While Cooperating
OpenAI’s response to the government’s request is a study in corporate diplomacy. The company complied fully. It shared its release plans and model capabilities with the government in advance, vetted its 20 preview partners with official input, and launched in the restricted format as requested.
However, OpenAI also publicly stated its displeasure. In the same blog post announcing the limited preview, the company said it does not believe “this kind of government access process should become the long-term default.” CEO Sam Altman told employees directly that a restricted, customer-by-customer government approval model is “not our preferred long-term model.” The company expressed specific concern that the arrangement blocks access for developers, cybersecurity professionals, international partners, and researchers who have legitimate uses for GPT-5.6.
“It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them.”
— OpenAI statement on the limited preview arrangement
Dean Ball, a former White House AI adviser and incoming OpenAI employee, went further. He argued that Trump’s executive order has created what is effectively a de facto involuntary licensing regime for frontier AI, despite the order’s explicit prohibition on mandatory preclearance. The danger, Ball argued, is that without clearly defined safety standards, the government could delay launches indefinitely handing an advantage to Chinese AI developers who face no such restrictions and are shipping capable open-weight models like GLM-5.2 on their own timelines.
Frequently Asked Questions
1. Why did OpenAI delay the GPT-5.6 public rollout?
OpenAI delayed the full public launch at the request of the White House’s Office of the National Cyber Director and the Office of Science and Technology Policy. The government asked that access be limited to roughly 20 vetted organizations while a formal frontier model review framework is developed under Trump’s June 2, 2026 executive order.
2. What are GPT-5.6 Sol, Terra, and Luna?
GPT-5.6 is a three-tier model family. Sol is OpenAI’s most powerful model to date, targeting advanced reasoning, coding, and cybersecurity. Terra offers GPT-5.5-level performance at half the cost, making it the balanced enterprise default. Luna is the fastest and cheapest tier, suited for high-volume summarization, drafting, and automation tasks.
3. What is GPT-5.6 Sol pricing?
Sol costs $5 per million input tokens and $30 per million output tokens. Terra is $2.50 input and $15 output. Luna is $1 input and $6 output per million tokens. OpenAI also improved prompt caching, with cache reads discounted by 90% and cache writes billed at 1.25× the uncached input rate.
4. When will GPT-5.6 be available publicly?
OpenAI has said it plans to make GPT-5.6 Sol, Terra, and Luna generally available via ChatGPT, Codex, and the API in “the coming weeks.” The government has expressed support for a broader rollout, pending any concerns from the additional testing period. No fixed date has been announced.
5. Does the US government control who can use GPT-5.6?
During the current limited preview, yes effectively. The roughly 20 companies with access are those whose participation OpenAI shared with the government. Trump’s executive order technically makes this voluntary, and the order explicitly bars the creation of a mandatory licensing regime. However, OpenAI and observers have noted that the arrangement functions like one in practice.
6. What happened with Anthropic’s models and the US government?
Earlier in June 2026, the US government issued an export control directive against Anthropic over national security concerns tied to jailbreaks in Claude Fable 5. Anthropic suspended access to both Fable 5 and Mythos 5. As of late June 2026, Mythos 5 access has been partially restored for a limited group after Anthropic addressed the government’s concerns. Fable 5 remains restricted.