OpenAI began rolling out its GPT-5.6 model family to the general public on July 9, ending nearly two weeks in which access was limited to a small group of vetted partners under a U.S. government review process.

What changed

The three models — Sol, Terra and Luna — were first previewed on June 26 but withheld from broad release after the Trump administration’s early-June executive order on AI cybersecurity, which requires companies to submit sufficiently powerful models for government review 30 days before public launch. The Commerce Department’s Center for AI Standards and Innovation tested the models, and OpenAI said it sent technical staff to Washington to address officials’ concerns before receiving approval to proceed. OpenAI said it does not believe this kind of government access process “should become the long-term default,” but complied to move toward release.

Capabilities and pricing

Sol is OpenAI’s new flagship model, which the company says sets a state-of-the-art result on the Terminal-Bench 2.1 coding benchmark and posts competitive scores on the ExploitBench cybersecurity evaluation while using roughly a third of the output tokens of rival systems. On OpenAI’s SecureBio biology evaluations, Sol scored about nine percentage points above GPT-5.5, including 53.5% on virology questions and 68.4% on human-pathogen questions.

Terra is pitched as matching GPT-5.5’s performance at half the cost, and Luna is the cheapest of the three. According to OpenAI’s published API pricing, Sol costs $5 per million input tokens and $30 per million output tokens, Terra costs $2.50/$15, and Luna costs $1/$6. The release also adds a “max” reasoning-effort setting and an “Ultra mode” that coordinates subagents on complex tasks.

The rollout is reaching ChatGPT, Codex and the API simultaneously and is expected to reach full global availability within about 24 hours, OpenAI said.

Read also