OpenAI previews GPT-5.6 Sol, Terra, Luna

OpenAI opened a limited preview of its GPT-5.6 model family on June 26. It comes in three tiers: Sol is the flagship, Terra is the mid-range pick for everyday tasks, and Luna is the fast, low-cost one. The same release adds two new reasoning controls. A max effort level gives Sol the most time to think on hard problems, and an ultra mode goes past a single agent by spinning up subagents to split complex work.

On coding, OpenAI says GPT-5.6 Sol in ultra mode set a new state of the art on Terminal-Bench 2.1, the test for command-line tasks that need planning, iteration, and tool use, with a score of 91.91. As with every vendor leaderboard number, that score is self-reported, so treat it as a claim to verify rather than a settled fact.

The other two tiers are about cost. OpenAI says Terra matches GPT-5.5 while being roughly half the price, and Luna is its cheapest capable model yet.

What it means for you

You cannot use any of this today. The preview is limited to about 20 companies whose access the government approved, with a wider release promised in the coming weeks. That gating is the same story behind the White House asking OpenAI to slow-roll the launch. When the models do open up, decide on real work, not the headline number: run the verification habit on your own tasks, the way you would weigh Gemini’s computer-use benchmarks before trusting them.