GPT-6 Astra vs Sol vs Luna: Which One Runs Your ChatGPT and Your Dot? (2026)
Quick answer: GPT-6 Astra is OpenAI's flagship and the one your Dot runs on; it's also the default in paid ChatGPT since September 25. GPT-6 Sol is the cheaper everyday model, GPT-6.1 Sol (announced September 29) gets close to Astra at Sol's price, and GPT-6 Luna is the fast, low-cost model free users get in the desktop app. For most people the choice is made for them by their plan. And for everyday admin like reminders, follow-ups and group messages, which model is underneath matters far less than whether the assistant is where you actually talk, which for many people is WhatsApp.

Your AI assistant, right in WhatsApp.
Reminders that fire on time, follow-ups handled for you, PDFs summarized in seconds, and your Google Calendar run by text. Nothing to install. Just message Sonic like any contact.
Four models in four weeks
September was a busy month for OpenAI's naming team. GPT-6 Astra arrived on September 3. GPT-6 Sol and GPT-6 Luna followed on September 22. Then at DevDay on September 29, OpenAI announced GPT-6.1 Sol and launched dots, its always-on personal agents.
If you're confused about which one you're using, you're not alone. The good news: for most people, the answer is decided by your plan and the product you open, not by a setting you need to find.
This guide covers what each model is for, which one sits behind ChatGPT and behind a Dot, what each costs, and the part that gets lost in the launch posts: when the model genuinely doesn't matter.
| Model | Released | What it's for | Where it runs |
|---|---|---|---|
| GPT-6 Astra | Sept 3 | Hardest, longest work; computer use | Paid ChatGPT (default), OpenAI Dots, API |
| GPT-6.1 Sol | Sept 29 | Near-Astra quality at Sol's price | API, Codex |
| GPT-6 Sol | Sept 22 | Everyday coding and professional work | ChatGPT Work, Codex, API |
| GPT-6 Luna | Sept 22 | Fast, cheap, high-volume tasks | Free and Go (desktop app), ChatGPT Work, Codex, API |
GPT-6 Astra: the flagship (and your Dot's brain)
Astra is OpenAI's most capable model. OpenAI built it on its largest training run to date and pitches it at long, multi-step professional work: using a computer, browsing, research, coding. It became the default engine for paid ChatGPT plans on September 25.
It's also what powers OpenAI Dots. OpenAI's launch materials describe dots as always-on agents powered by GPT-6 Astra, each with its own cloud computer. So if you're asking 'which GPT-6 is my Dot using?', the answer is Astra.
Why Astra for agents? Because agents run unattended for a long time. OpenAI says Astra finishes a standard computer-use benchmark in roughly 40 minutes per task where the previous generation needed around 75, and scores higher. It's also better at staying inside the task you gave it, which matters when nobody is watching.
The catch is cost. Astra is priced as a flagship, several times Sol's per-token rate on the API. Inside ChatGPT you don't see that directly; it's part of why Dots start with Pro and Business Premium subscribers.
GPT-6 Sol and GPT-6.1 Sol: the workhorses
GPT-6 Sol arrived on September 22 as the everyday model: $2 per million input tokens and $10 per million output on the API, roughly half what GPT-5.6 Sol cost. OpenAI says it makes about half as many mistakes as its predecessor and it improves on coding and computer use.
One week later, GPT-6.1 Sol raised the bar again at the same price. OpenAI says it approaches Astra on coding, professional work and computer use at around one-fifth of Astra's token price. For developers and businesses paying per token, that's the headline of DevDay.
Sol isn't in ChatGPT's normal chat mode; it runs in ChatGPT Work, Codex and the API. If you're a regular ChatGPT user, you won't usually pick it. If you build products on OpenAI, it's likely where most of your traffic should go.
GPT-6 Luna: fast and cheap
Luna is the small one: $0.10 per million input tokens and $0.50 output on the API, about half of GPT-5.6 Luna. It's built for quick, repetitive, high-volume tasks like sorting, extracting and short replies.
It's also what free and Go ChatGPT users get in the desktop app. That's worth knowing: if you're on the free plan and ChatGPT feels less sharp on complex requests than a friend's paid account, this is why.
Luna isn't a bad model. It's the right model for jobs that don't need deep reasoning, and it answers quickly.
So which one should you care about?
If you're on ChatGPT Plus or higher: you're already on Astra for normal chats. Nothing to change.
If you're on ChatGPT Pro or Business Premium and use a Dot: it's Astra. OpenAI hasn't announced a way to run a Dot on a cheaper model.
If you're on the free plan: you're on Luna in the desktop app. Fine for quick tasks; upgrade if you need heavy reasoning.
If you're building on the API: Sol or 6.1 Sol for most work, Luna for volume, Astra for the hardest long-running agent tasks.
And if your question is really 'which AI should handle my admin?', the model is the wrong place to start.

Your AI assistant, right in WhatsApp.
Reminders that fire on time, follow-ups handled for you, PDFs summarized in seconds, and your Google Calendar run by text. Nothing to install. Just message Sonic like any contact.
When the model matters less than where it lives
Here's the thing nobody's launch post says. For a lot of everyday admin, the gap between Astra and Luna is irrelevant.
"Remind me to renew the car insurance on the 3rd." "Tell the family group the dinner moved to 8." "What did the electrician quote?" "What's the payment date on this invoice?" None of these needs a frontier model on a cloud computer. They need an assistant that fires on time, remembers what you said, and is sitting in the conversation you already have open.
That's where every GPT-6 model has the same limit: none of them is in WhatsApp. Astra powers Dots, which live in ChatGPT, Slack and Teams. Sol and Luna live in ChatGPT, Codex and the API. None of them can be messaged from a WhatsApp chat or added to a WhatsApp group.
For people in WhatsApp-first countries (Israel, most of Europe, Latin America, India), that's where the customers, suppliers, team and family already are. An assistant somewhere else is one more app to remember to open.
The assistant that's already in your WhatsApp
Agent Sonic is a personal AI assistant that lives entirely inside WhatsApp. There's no app to install and no dashboard. You message it like any contact.
Reminders, one-off and recurring: "Remind me to pay the VAT on the 15th every other month." Group reminders that post on their own: "Every Thursday at 6, remind the team group to submit hours." A memory you can query: "What did I agree with the landlord about the parking spot?"
Documents and voice, in the chat: forward a contract and ask "what's the notice period?", forward a spreadsheet and ask "total by client." Send a voice note while driving and get a reply as text or voice.
Follow-ups on your behalf: "Ask the three caterers I saved for a quote for 40 people on the 20th and tell me who answers." Sonic messages them and reports back. It also browses the web for live answers, creates and edits images, and connects to Google Calendar, Sheets and Docs.
What Sonic isn't: a computer-use agent. It won't log into websites and click through forms. If you need that, a Dot on Astra is the right tool. For the everyday admin that happens in WhatsApp, Sonic does it where the chat already is.
Agent Sonic is available now. Leave your details at https://www.tryagentsonic.com/contact to get Sonic's number and onboarding; pricing is on the site.
Bottom line
Astra is the flagship and your Dot's brain. 6.1 Sol is the new value pick for developers. Sol is the everyday workhorse. Luna is fast, cheap, and what free users get. Your plan mostly picks for you.
If you're choosing an assistant for the admin that fills your day, start with where you talk, not which model is underneath. If that's WhatsApp, Agent Sonic is already there.
Frequently asked questions
Which GPT-6 model does my OpenAI Dot use?
GPT-6 Astra. OpenAI describes dots as always-on agents powered by GPT-6 Astra. The cheaper GPT-6 models are aimed at ChatGPT Work, Codex and the API, and OpenAI hasn't announced a way to run a dot on them.
Which model does ChatGPT use by default now?
On a paid plan, normal ChatGPT chats run on GPT-6 Astra, which became the default on September 25, 2026. Free and Go users get GPT-6 Luna in the desktop app.
What's the difference between GPT-6 Astra, Sol and Luna?
Astra is the flagship: most capable, slowest, most expensive. Sol is the everyday workhorse at roughly one-fifth of Astra's price. Luna is the fast, cheap model for simple, high-volume tasks. GPT-6.1 Sol, announced September 29, gets close to Astra on professional work at Sol's price.
How much do the GPT-6 models cost?
Through the API, GPT-6 Sol and 6.1 Sol are $2 per million input tokens and $10 per million output. Luna is $0.10 and $0.50. Astra is priced as the flagship at several times Sol's rate. Inside ChatGPT, you pay your plan price rather than per token.
Can any GPT-6 model run in WhatsApp?
No. None of the GPT-6 models, and none of OpenAI's agents, work inside WhatsApp. Agent Sonic is built to handle reminders, group messages, follow-ups and documents there.

Your AI assistant, right in WhatsApp.
Reminders that fire on time, follow-ups handled for you, PDFs summarized in seconds, and your Google Calendar run by text. Nothing to install. Just message Sonic like any contact.