If you mainly use chat-style AI and you're only now trying to keep up with new tools, this is the part that can waste time, money, and attention fast. You see The new GPT-5.6 family: Luna, Terra, Sol, think it's just another release post, and keep scrolling. But the line that actually changes your next move is simpler: if you don't name the sub-model, you're effectively buying Sol.
That matters because most people read gpt-5.6 as the safe, neutral default. It isn't. In the official model guide, the plain gpt-5.6 alias routes to gpt-5.6-sol, while Terra and Luna have to be requested explicitly [S002]. In normal words: the short name is not just shorthand. It already makes the speed, quality, and cost choice for you.
The pricing is why this stops being a naming detail and turns into a real decision. As of July 9, 2026, Sol is priced at $5 input / $30 output per 1M tokens. Terra is $2.50 / $15. Luna is $1 / $6 [S001]. You do not need to be an engineer to care about this. If you paste gpt-5.6 into a tool, workflow, or API setting and assume you can think about cost later, you may already be in the highest-priced lane before you notice.
This is not a cheap-is-always-better take. Sol may be the right default for higher-value work, and that part gets lost when people flatten the whole family into one name. A model update is worth your attention only when it changes your next decision. Here, the decision is whether you want the family name or a specific tier.
So the sentence worth saving is this: if you don't name the sub-model, you're effectively buying Sol. If you know someone who is about to test GPT-5.6 by typing the family name and assuming Luna or Terra will show up automatically, share this with them. It is a small naming detail, but it changes what you get and what you pay.