- Published on
OpenAI Rolls Out GPT-5.6 But Only for Some Users Due to Trump Admin
- Authors
- Name
- DeepHorus
OpenAI on Friday unveiled the GPT-5.6 family of AI models, launching a limited preview of its new models codenamed Sol, Terra, and Luna before a broader planned release in the coming weeks. The restricted rollout follows a request from the Trump administration while officials develop a framework for evaluating advanced AI systems, confirming earlier reports. This marks the second time this month the White House has impacted a frontier AI model's release, following a similar request to Anthropic regarding its Fable 5 and Mythos 5 models.
OpenAI said the limited preview is part of ongoing engagement with the U.S. government. The company previewed its plans and the models’ capabilities ahead of launch, and at the administration's request, it is starting with a small group of trusted partners whose participation has been shared with the government. The company emphasized that the release will expand more broadly once a process for future frontier AI releases is established. This move highlights the growing federal oversight of cutting-edge AI technologies under President Donald Trump's administration.
GPT-5.6 includes three model variants: Sol, the flagship model; Terra, a balanced model for everyday work; and Luna, a fast and affordable option. OpenAI stated that Sol improves performance across coding, biology, and cybersecurity, while Terra delivers performance comparable to GPT-5.5 at half the cost, and Luna brings strong capability at the lowest cost. The release also introduces new reasoning modes—"max" and "ultra"—that allow Sol to allocate more time for complex problems or coordinate multiple subagents for demanding tasks.
In testing, OpenAI claimed Sol achieved the highest score on TerminalBench, a benchmark for command-line software engineering tasks, outperforming GPT-5.5, Claude Mythos 5, and Fable 5. On GeneBench v1, which evaluates long-horizon genomics and quantitative-biology analyses, Sol achieved stronger results than GPT-5.5 while using fewer tokens. The company also highlighted Sol's cybersecurity capabilities, noting it pairs stronger defensive security research support with expanded safeguards against offensive misuse, though it remains below OpenAI's Cyber Critical threshold.
OpenAI explained that although GPT-5.6 Sol can identify vulnerabilities and exploit components, it could not autonomously produce a complete exploit chain during testing. The model is trained to refuse prohibited cyber assistance, including when users attempt to disguise intent or jailbreak the model. These safeguards establish a first boundary around acceptable use, the company said. This cautious approach aligns with the administration's focus on responsible AI development.
During the preview, GPT-5.6 Sol, Terra, and Luna will be available through the API and Codex to select partners before expanding publicly to ChatGPT and other users. OpenAI also introduced a new naming system for the model family and announced that GPT-5.6 Sol will launch on Cerebras in July, offering inference speeds of up to 750 tokens per second. Broader availability is expected in the coming weeks as the evaluation process with the government progresses.
The industry is watching closely how this federal oversight shapes future AI rollouts. The Trump administration's active role in evaluating frontier models could set a precedent for more structured pre-release testing requirements. With GPT-5.6's improved reasoning and cybersecurity features, the model represents a significant step forward, but its phased release underscores the balancing act between innovation and regulation in the AI sector.
