OpenAI releases GPT-5.6 in limited preview pending US goverment sign off

OpenAI released three versions of GPT-5.6 in a limited preview, but the big question is whether the US government allows the model to graduate to broad release.

The GPT-5.6 models--Sol, Terra and Luna--are the first frontier models to launch since the US government sidelined Anthropic's Fable 5 and Mythos 5. Anthropic launched its latest models and had to pull them back. Negotiations with the government are ongoing.

• Anthropic’s Fable 5 and Mythos 5 dispute: Still a mess
• Anthropic shuts down Fable 5, Mythos 5 as US govt shuts down foreign national access
• Anthropic’s Dario Amodei wants governments to have the power to block ‘dangerous’ AI systems
• Anthropic’s Claude Fable 5, Mythos 5: What you need to know

Enter OpenAI's launch where perhaps the most interesting thing in the release is what the company said about the launch process with the government. OpenAI said:

"GPT‑5.6 Sol launches with our most robust safety stack to date. We strengthened protections for higher-risk activity, sensitive cyber requests, and repeated misuse, and spent multiple weeks finding weaknesses, pressure-testing our system, and hardening it against real-world attacks.

We believe in broad access, and we plan to make GPT‑5.6 Sol, Terra, and Luna generally available in the coming weeks. As part of our ongoing engagement with the U.S. government, we previewed our plans and the models’ capabilities ahead of today’s launch. At their request, we are starting with a limited preview for a small group of trusted partners whose participation has been shared with the government, before releasing more broadly. During this preview, we will continue testing and coordinating closely with partners as we work toward broader availability.

We don’t believe this kind of government access process should become the long-term default. 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."

In other words, the exact launch process for frontier models is a work in progress, but the status quo could hurt innovation.

Here's what you need to know about GPT-5.6 assuming it reaches broad distribution.

• GPT-5.6 Sol is OpenAI's strongest model yet and excels in coding, biology and cybersecurity. GPT 5.6 Sol has a max reasoning setting and an ultra mode that leverages subagents.
• OpenAI said GPT-5.6 Sol is competitive with Mythos Preview on cybersecurity and can conduct vulnerability and exploitation research.
• The company emphasized that GPT-5.6 models use fewer output tokens in a nod to the rather obvious backlash over AI costs.
• GPT-5.6 Terra is a balanced model for daily tasks with Luna aiming for speed and affordability. OpenAI said Terra is competitive with GPT-5.5 and 2x cheaper.
• OpenAI said GPT-5.6 Sol is better at finding and fixing vulnerabilities than attacks. OpenAI is walking the line between touting cybersecurity capabilities without raising alarm bells.
• GPT-5.6 Sol will launch on Cerebras at speeds of up to 750 tokens per second in July.

As for the pricing, OpenAI said:

"GPT‑5.6 is priced per 1M tokens across three model sizes: Sol is $5 input / $30 output; Terra is $2.50 input / $15 output; and Luna is $1 input / $6 output. GPT‑5.6 also introduces more predictable prompt caching, including support for explicit cache breakpoints and a 30-minute minimum cache life. For GPT‑5.6 and later models, cache writes are billed at 1.25x the model’s uncached input rate, while cache reads continue to receive the 90% cached-input discount."