OpenAI just released three new models with names instead of numbers, and if you run a business you do not need to understand the benchmarks. You need to know which one to reach for and when. This is the plain-English version, with a common finance task for each so you can feel exactly where each one fits.
On July 9, 2026, OpenAI released its GPT-5.6 family, and instead of one model it comes in three tiers with actual names: Sol, Terra, and Luna. The number tells you the generation. The names tell you the horsepower. Sol is the flagship, Terra is the balanced middle, and Luna is the fast, affordable one.
Here is the part that matters for you. More power is not always better. A bigger, smarter model costs more and takes longer, so pointing your most expensive model at a task that sorts receipts is like hiring a senior strategist to alphabetize a filing cabinet. The whole skill is matching the model to the job. Once you see it through that lens, the choice gets simple.
The cleanest way to feel the difference is to run the same part of your business through all three: your finances. Every business has money tasks, they range from mindless to high-stakes, and that range lines up almost perfectly with these three models. Let's walk through it.
Before you pick a model, ask yourself one thing about the task in front of you. If this goes wrong, what does it cost me? A miscategorized receipt costs you a two-second fix. A wrong number in a funding decision costs you thousands, or the decision itself. The higher the cost of being wrong, the more model you want. Hold that question in your head as we go through the three.
Luna is the fastest and cheapest of the three. It is built for tasks you do in bulk that do not need deep judgment, where speed and low cost matter more than raw brainpower.
The finance task: categorizing expenses. Say you have a spreadsheet of 200 transactions that need to be sorted into buckets like software, travel, contractors, and meals before they go to your bookkeeper. There is no strategy here, just a lot of small, repetitive calls. That is Luna's sweet spot. If it gets one wrong, you fix it in seconds, so paying flagship prices for it would be a waste.
Here is a list of business transactions with descriptions and amounts. Sort each one into a category from this list: software, travel, contractors, meals, office, marketing, other. Return it as a clean table with the original description, the amount, and the category you assigned. Flag anything you are unsure about in a separate column so I can review it. Here is the list: [paste your transactions]
Terra is the balanced middle tier, and for most founders it is the one you will live in. It is smart enough for real analysis and writing, without the cost or the wait of the flagship. If you are not sure which to use, this is your default.
The finance task: your monthly report. Every month you have raw numbers, revenue, expenses, cash in the bank, and someone has to turn them into a clear summary that tells you how the business is actually doing. That takes real reasoning and clean writing, but it is standard work you repeat every month. That is Terra exactly. It handles the thinking and the polish, and it does it quickly.
You are my finance summarizer. Below are this month's numbers: revenue, expenses by category, and cash on hand, along with last month's totals for comparison. Write a one-page monthly financial summary a busy founder can read in two minutes. Include what changed versus last month, the two or three things worth my attention, and one plain-English sentence on whether cash flow is healthy. Here are the numbers: [paste your numbers]
Sol is the flagship, the smartest and the most expensive. You reach for it when the task is complex, the reasoning is heavy, and the cost of being wrong is real. This is not your everyday model, it is the one you bring in for the big calls.
The finance task: a hire-or-wait decision. You are deciding whether you can afford to bring on your first full-time employee. That means projecting cash flow across a few different scenarios, best case, worst case, and likely, and thinking through what happens to your runway in each. The numbers interact, the stakes are high, and a shallow answer could sink you. That is where Sol earns its price. It can hold the whole messy picture and reason through it carefully.
Act as a careful financial analyst. I am deciding whether to hire my first full-time employee at a salary of [amount] plus roughly 20 percent for taxes and benefits. Here is my current monthly revenue, my average monthly expenses, and my cash on hand: [paste the numbers]. Build three cash-flow scenarios over the next 12 months: best case, likely, and worst case, each with a reasonable assumption about revenue growth. For each, show my projected cash balance month by month and tell me in which month, if any, I run out of runway. Then give me your honest read on whether this hire is affordable now, and what would need to be true for it to be safe.
For a decision this size you can also use Sol's higher-effort settings, which OpenAI offers for its most demanding work. The point is not the setting name. The point is that you save the heavy model for the heavy decisions.
Here is the whole thing on one screen. When a task lands on your desk, ask what it would cost you to get it wrong, then pick accordingly.
| Model | Best for | The finance example |
|---|---|---|
| Luna | High-volume, low-stakes, repetitive work where speed and cost matter most | Sorting 200 expenses into categories for your bookkeeper |
| Terra | Everyday work that needs real thinking but is standard and repeatable. Your default | Turning this month's numbers into a clear one-page report |
| Sol | Complex, high-stakes decisions where being wrong is expensive | Modeling three cash-flow scenarios for a big hire |
One more mindset shift, because it changes how you budget for this. AI models are like streaming services. No single one carries every show, so most households pay for a few: one for this, one for that. AI works the same way. OpenAI is strong at some things, Claude at others, Gemini at others, and a serious business often keeps more than one subscription to get the best of each.
Sol, Terra, and Luna are the tiers inside one of those services. Picking between them is like choosing the right plan within a single subscription. Choosing whether to also run Claude or Gemini alongside OpenAI is like deciding whether Netflix alone is enough or you also want Max. There is no single right answer, only the right mix for the work you actually do. Start with one, get fluent, and add another when a real need shows up.
You do not have to memorize a single benchmark. You just have to ask what a mistake would cost, then match the model to the moment: Luna for the mindless volume, Terra for the everyday work, Sol for the decisions that matter. Do that and you will spend less, move faster, and stop overthinking which button to press.
When you are ready to turn this into a real system for your whole business, deciding which model runs which task, and building the workflows around them, that is exactly what we do inside Her AI Systems™.
They are the three models in OpenAI's new GPT-5.6 family, released July 9, 2026. Sol is the flagship built for the hardest, highest-stakes work. Terra is the balanced everyday model. Luna is the fastest and most affordable. The number is the generation, and the names are durable tiers that will carry forward.
Terra. It is the balanced middle tier built for the standard work you do every week, like turning your raw numbers into a clean monthly report. Reach up to Sol only when the stakes and complexity are high, and drop down to Luna for high-volume, low-judgment tasks.
No. On a paid ChatGPT plan you get access to all three and simply pick the one that fits the task. Free and Go users get Terra. You are choosing a setting inside one subscription, not buying three separate products.
Often, yes. Think of AI models like streaming services. No single one has everything, so serious businesses often keep more than one subscription to get the strengths of each. OpenAI, Claude, and Gemini each shine at different work, and picking a tier inside OpenAI is like choosing the right plan within one service.
For most founders the practical answer is that Sol is the premium tier, Terra is the mid tier, and Luna is the budget tier. On the API, priced per million tokens, Sol is $5 in and $30 out, Terra is $2.50 in and $15 out, and Luna is $1 in and $6 out.
Free Download
Keep the model cheat sheet close for the next time you are not sure which one to open.
Follow along for the next guide
New free guide every Tuesday. Find them first on Instagram, TikTok, or by subscribing to the email list.
Want more?
One new free guide. Every Tuesday. No spam, no fluff, just the systems and prompts I'm building for clients, simplified for you.
Subscribe