OpenAI rolls out ChatGPT 5.6 models after US government review

OpenAI has released its GPT-5.6 family after the US government completed a security review that had impeded the rollout of the model. The Trump administration had requested federal agencies to review the model for potential security risks.

Enterprise and individual users who purchase AI by the token will find this news relevant, as OpenAI is marketing this model on its cost efficiency rather than raw intelligence. 

Sam Altman, CEO of OpenAI, told CNBC that the flagship version is 54% more efficient on agentic coding tasks than its competitors.

Three versions were released after a delayed launch

GPT-5.6 came in three different versions: Sol, Terra, and Luna. Sol focuses on complex reasoning and deep coding tasks. Terra is best for your everyday, run-of-the-mill tasks, and Luna, being the cheapest, is perfect for high-volume tasks. 

Getting to Thursday took a detour through Washington. Access had been restricted to about 20 government-vetted partners while the Commerce Department’s Center for AI Standards and Innovation ran its evaluation, with OpenAI sending technical staff to answer the agency’s questions. 

Altman said the company “made many changes” to the system during the review, which he characterized as “a collaborative back and forth.” The process pulled in Commerce Secretary Howard Lutnick, Treasury Secretary Scott Bessent, and National Cyber Director Sean Cairncross.

Why was the US government involved in the GPT-5.6 launch?

This has to do with the immense coding abilities of some AI models. OpenAI’s GPT-5.6 and Anthropic’s Mythos series can spot software weakness. This has understandably set off alarm bells among government officials who are worried these advanced models could be used to target financial institutions running on older AI models.

This particular review went beyond the voluntary review framework signed by President Trump on June 2. The framework asks companies to allow government agencies to inspect strong AI models before they’re released. 

The hoops OpenAI had to jump through are not much different from Anthropic’s experience with the US government. Fable 5 and Mythos 5 models were initially paused by the government over security concerns. But Fable 5 has been fully restored, while Mythos 5 is available only to a few US organizations.

OpenAI ships new AI agent with new model

Alongside the launch of GPT-5.6, OpenAI launched ChatGPT Work, an AI agent built on GPT-5.6 that can operate across a user’s apps and files, if connected. The agent can also assemble documents, spreadsheets, and presentations, and is available on web, phone, and desktop. It is an autonomous user that runs for hours on a single prompt.

Comparisons with Anthropic’s models have already begun, with some labeling Fable as possessing raw intelligence while GPT-5.6 works best for daily tasks. Dan Shipper, CEO of Every, put it in a post on X: “If you need to get across the galaxy, use Fable. If you need to get around town using the best available tool for the job, use 5.6.”

The voluntary framework signed by President Trump is one to watch. Because the government’s actions will go a long way in determining security restrictions.

If you're reading this, you’re already ahead. Stay there with our newsletter.