A Step From "Preview" Toward "Available"
The GPT-5.6 that OpenAI previewed on June 26 comes in three tiers—Sol (hard-problem flagship), Terra (everyday balanced), and Luna (fast, high-volume). Heading into July, the rollout cadence began to diverge: the balanced tier Terra reportedly opened an enterprise API waitlist first, the fast tier Luna is in small-scale testing for some developers, while the strongest Sol remains under Washington's "frontier-model review" gating with no public timeline. Not all of the above tier progress has been confirmed point by point officially; the specifics are per OpenAI News disclosures.
Why the Middle Tiers Go First
Releasing Terra and Luna first while keeping Sol behind the gate follows a real logic: the stronger the model, the more regulatory and safety-review checkpoints. The middle tiers can meet the everyday needs of the vast majority of enterprises and developers without triggering the strictest review, so they become the "get it working first" tier. For most teams, whether you can get Sol isn't actually the pressing matter.
What It Means for Selection
The signal of this step isn't how strong the model is, but that "availability" is becoming a hard constraint in selection. Within the same model family, different tiers have different opening times, regions, and compliance thresholds—so in procurement and technology selection, besides capability and price, you have to factor in "which tier can I get, when can I get it, and is it compliant." The strongest tier sits in the display window while the key is in the reviewers' hands—this is likely to be the norm for a while.
via: OpenAI News