GPT-5.6 (Sol, Terra & Luna): What OpenAI's New Model Family Means for Your Business
By Jonathan Miksis · Updated July 8, 2026 · 9 min read
Most founders have been overpaying for AI without realizing it.
Not because the tools are bad. Because they have been using the most expensive model for work that never needed the smartest model in the room.
That is why GPT-5.6 matters.
OpenAI did not just release a smarter model. It split the new model family into three clear tiers: Sol, Terra, and Luna. Each one is built for a different kind of work, with different speed, capability, and cost.
If you run a business, that matters more than any benchmark. It means you stop asking "is the new model better?" and start asking the question that actually saves money: "which tier should I use for this job?"
I have spent the last two years using AI as infrastructure across multiple businesses. Every model release, I ask the same thing: what changes for my Monday morning? Here is the founder translation of GPT-5.6, minus the leaderboard drama.
One honest note up front: OpenAI previewed GPT-5.6 in late June, and its developer announcement points to July 9 as the next availability milestone for Sol, Terra, and Luna. The rollout is still phased. Preview access starts with developers, enterprises, and selected trusted partners through the API and Codex, with broader availability across ChatGPT, Codex, and the API expected to arrive in stages. So depending on your plan and account type, you may not see all three options immediately. This is the founder playbook for when they land in your account.
The TL;DR for busy founders
- Sol = the flagship specialist. For your hardest problems: complex analysis, deep coding, security. Most capable, most expensive.
- Terra = the everyday operator. Built for high-volume business work (support, internal tools, document analysis) and about 2x cheaper than GPT-5.5 for the same tasks.
- Luna = the fast assistant. For quick drafts, summaries, and routine automation. Cheapest and fastest.
- The real shift: OpenAI now separates the model generation from the job it is meant to do. GPT-5.6 is the generation. Sol, Terra, and Luna are the workers. You pick the worker that fits the task instead of running everything through the most expensive model.
Why "three tiers" is actually good news for non-technical people
For two years, the AI conversation has been "which single model is best?" That framing never fit how a business actually works. You don't hire one person to do everything, you match the person to the task.
GPT-5.6 finally works the same way. Think of it like staffing:
- Sol is your senior specialist. You don't put them on routine email. You save them for the pitch deck analysis, the messy financial model, the strategic call that needs real reasoning.
- Terra is your reliable operator. The one who handles the bulk of the business: answering customers, processing documents, running the internal tools. Dependable, and now noticeably cheaper.
- Luna is your fast assistant. Quick drafts, summaries, "clean this up," the thousand small tasks that shouldn't cost much or take long.
Once you see it this way, the cost math gets obvious. Most founders were paying flagship prices to do assistant-level work. GPT-5.6 lets you stop doing that. For most small businesses, the biggest win is not using Sol more. It is using Sol less.
Which GPT-5.6 tier for which job?
Here's how I'd map it to the work founders and creators actually do:
| Your task | Use | Why |
|---|---|---|
| Analyze a competitor, pressure-test strategy, model finances | Sol | Hard reasoning is worth the premium |
| Draft & debug an automation or a real technical build | Sol | Top coding tier |
| Answer customer emails, summarize call notes, tag support tickets | Terra | High volume, now ~2x cheaper |
| Analyze a stack of documents or reports | Terra | Balanced power for bulk knowledge work |
| Rewrite a subject line, summarize an article, first-draft a post | Luna | Fast and cheap wins here |
| Batch-generate social captions or meta descriptions | Luna | Routine volume, lowest cost |
Here is what this looks like in a real small business workflow.
A founder could use Luna to summarize customer feedback, Terra to group the complaints into patterns, and Sol to decide which product change actually matters. Same workflow, three levels of AI, much better cost control.
That is the new skill. Not just writing better prompts. Routing the right job to the right model.
What each tier costs
Pricing is per 1M tokens (roughly 750,000 words in, or out). For everyday use inside ChatGPT you won't count tokens, but if you run automations through the API, this is the math that matters:
| Tier | Input | Output | Best for |
|---|---|---|---|
| Sol | $5 | $30 | Hardest problems, coding, deep analysis |
| Terra | $2.50 | $15 | High-volume business tasks (2x cheaper than GPT-5.5) |
| Luna | $1 | $6 | Fast, routine, everyday work |
The Terra number is the sleeper. If you're running any kind of AI workflow at volume, doing the same work for half the cost of GPT-5.5 drops straight to your bottom line. Terra is the tier I would watch most closely. Sol will get the attention because it is the flagship. Luna will be useful because it is cheap. But Terra is where a lot of real business automation probably lives: good enough for serious work, cheap enough to run often.
The bigger shift: AI as a team, not a chatbot
The feature pattern I would watch most closely is parallel agent work. Instead of one model working through a request step by step, the frontier is moving toward systems that split work across multiple agents, run pieces in parallel, and bring the result back together. That is the real shift for founders.
If that sounds familiar, it's because it's happening across the whole frontier, not just at OpenAI: Anthropic's Claude did a version of it with agent teams (I broke that down in what Claude Opus means for founders). The mental model is the point. You stop thinking "what can I ask this AI?" and start thinking "what can I delegate to this team?" The founders who internalize that first will operate at a different speed.
You won't feel any of this as a benchmark. You'll feel it as a support queue that clears faster and a strategy session that actually holds the whole picture.
What to do this week
Don't overhaul anything. Do this:
- When GPT-5.6 lands in your account, pick one repeatable task you already run on AI, such as drafting customer replies, summarizing sales calls, repurposing blog posts, or analyzing support tickets. Move that task to the cheapest tier that still gives you a good result. Start with Luna or Terra before reaching for Sol.
- Reserve Sol for one hard thing. Feed it your messiest real problem: the quarter's numbers, the competitor you're worried about, the plan you haven't pressure-tested. Ask it to challenge your thinking, not summarize it.
- Right-size the rest. Anywhere you're using a flagship model for simple work, drop it to Luna. That's free money.
If you want the exact prompts I would use to test Sol, Terra, and Luna, grab my free AI prompt pack. And if you want help turning this into a real workflow for your business, that is what my done-for-you AI plans are built for.
GPT-5.6 vs Claude vs Gemini: the honest take
You don't have to pick one. My daily stack uses all three for what each does best, and GPT-5.6's tier system doesn't change that, it just makes ChatGPT cheaper to run at the right level. Claude still leads on long-context and agentic writing work; Gemini still wins on live research and anything inside Google. I keep the full breakdown of the AI tools I actually use updated as these ship, and I put the three families head to head in my ChatGPT vs Claude vs Gemini guide.
The winning move in 2026 is not betting your business on one model. It is knowing which tool, and now which tier, to reach for.
Frequently asked questions
What is GPT-5.6 Sol, Terra, and Luna? They're three tiers of OpenAI's GPT-5.6 generation. Sol is the flagship for the hardest problems (complex reasoning, coding, security), Terra is the balanced workhorse for high-volume business tasks at about half the cost of GPT-5.5, and Luna is the fast, cheap tier for everyday drafting and automation. The number (5.6) marks the generation; the names mark the job.
When can I use GPT-5.6? OpenAI's developer announcement points to July 9 as the next availability milestone for Sol, Terra, and Luna. Access is still rolling out in stages, starting with developers, enterprises, and selected trusted partners through the API and Codex. Broader access across ChatGPT, Codex, and the API is expected to expand over time, but availability may depend on your plan, workspace, and account type.
How much does GPT-5.6 cost? Per 1M tokens: Sol is $5 in / $30 out, Terra is $2.50 / $15, and Luna is $1 / $6. Inside a ChatGPT subscription you won't be counting tokens; the per-token pricing matters most if you run automations through the API.
Which GPT-5.6 tier should I use? Match the tier to the task: Sol for hard reasoning and coding, Terra for high-volume business work, Luna for fast routine drafting. The table above maps common founder tasks to each. When in doubt, start on the cheaper tier and only move up if the output isn't good enough.
Is GPT-5.6 better than GPT-5.5? Yes, and the bigger win for most businesses is cost: Terra does everyday work at roughly half the price of GPT-5.5. The clearest gains are on hard reasoning and coding work at the Sol tier.
GPT-5.6 or Claude, which should a founder use? Both, for different jobs. GPT-5.6's tiers make ChatGPT efficient to run at the right level; Claude Opus still leads on long-context and agent workflows. See my Claude Opus founder breakdown and my full AI stack.
Related reading
Get my free AI prompt pack
Join 300+ founders, creators, and professionals. I’ll send you my go-to AI tools and copy-paste prompts that save hours every week. No spam.
