OpenAI launched a limited preview of its GPT-5.6 model suite on Friday, comprising three models — Sol, Terra, and Luna — less than 24 hours after agreeing to delay the release at the request of the Trump administration.
OpenAI released a limited preview of its GPT-5.6 model suite on Friday, less than 24 hours after news emerged that the company had agreed to stagger the launch at the request of the Trump administration.
The new suite consists of three models: Sol, the flagship offering; Terra, a medium-tier model designed for "high-volume work"; and Luna, described as a "fast and affordable" everyday model. OpenAI said the suite is particularly capable at coding, cybersecurity, and biology, as well as sustaining focus during long-horizon agentic AI tasks.
Pricing for GPT-5.6 Sol is set at $5 per million input tokens and $30 per million output tokens — nearly half the cost of Anthropic's Claude Fable 5, which is priced at $10 input and $50 output, according to OpenAI. Terra is priced at half the cost of Sol, and Luna at less than half the cost of Terra.
OpenAI also introduced two additional modes for Sol: a "max" mode for deeper reasoning and an "ultra" mode for leveraging sub-agents.
The company said it dedicated "approximately 700,000 A100e GPU hours" to automated red-teaming ahead of the release, and that third-party testers will continue evaluating the models for the next two weeks.
Given an active security debate in Washington, OpenAI devoted the majority of its announcement to safety and potential misuse. The company wrote that "GPT-5.6 is trained to refuse prohibited cyber assistance, including when users attempt to disguise their intent or jailbreak the model." OpenAI also stated that Sol "is better at helping people find and fix vulnerabilities than reliably carrying out end-to-end attacks," and that it does not cross the cyber-critical threshold under the company's preparedness framework — a framework OpenAI revised in April, removing some areas of previous study.
The Trump administration is approving customers on a case-by-case basis during the preview period, reflecting the close governmental scrutiny surrounding the release. OpenAI acknowledged the arrangement but signaled it views the arrangement as temporary.
"We don't believe this kind of government access process should become the long-term default," the company wrote in its announcement. "It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them. 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 said the full suite should be generally available in the coming weeks, reflecting the company's stated commitment to "broad access."
The release comes as competitive pressure in the AI model market intensifies, with pricing and safety positioning emerging as central battlegrounds. Whether the administration's case-by-case review process extends to future model releases — and how OpenAI negotiates that framework — will be among the more consequential regulatory questions facing the company as it moves toward a wider rollout.
Marc Sabatini is a staff writer at TechEchelon covering enterprise software, cybersecurity, and the regulatory beats that shape both. He focuses on the deal flow and policy decisions that move markets.
More from Marc →Notion is shutting down Notion Mail on September 22, saying more than half of its users already manage email through AI agents without opening the inbox — a move that also closes the chapter on the Skiff email infrastructure Notion acquired in 2024.
Qualcomm unveiled its Dragonfly C1000 data center CPU with Meta as a launch customer, acquired AI software startup Modular for a reported ~$3.92 billion, and nearly doubled its 2029 non-handset revenue forecast to $40 billion, sending shares up 15%.
Nvidia CEO Jensen Huang said at the company's annual stockholder meeting that smuggled-chip AI data centers are a "dead end," while reaffirming that national security takes precedence over commercial opportunity in restricted markets like China.
Written for readers who already know the basics — markets, AI, and the policy decisions that shape both.