
ℹ️ Quick Answer: GPT-6 Sol and Luna are OpenAI’s two new smaller models, released September 22, 2026. They’re live in ChatGPT Work and Codex for paid users, and Luna is in the desktop app for free users. Neither one is in the regular ChatGPT chat yet. API pricing is half what GPT-5.6 cost.
📋 WHAT’S INSIDE
- Where GPT-6 Sol and Luna actually show up
- What the two models are for
- The price cut
- Where I land on it
- Frequently asked questions
Last updated September 23, 2026
OpenAI put out GPT-6 Sol and Luna yesterday, and the first thing I did was open ChatGPT to look for them. They weren’t there, and the announcement says as much.
I haven’t used either one yet(outside of 5.6) , so this is where they are, what they cost, and what OpenAI is claiming.
Where GPT-6 Sol and Luna actually show up
They’re in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users, and Luna is in the desktop app for Free and Go users. OpenAI’s announcement says outright that “these models are not yet available in Chat.”
Free and Go users: open the ChatGPT desktop app and look for GPT-6 Luna in the picker. If you use ChatGPT the normal way, typing into the chat window, nothing changed for you yesterday. The rollout is gradual, and the same page says to try again later if you don’t see them in Work or Codex. For developers, the API names are gpt-6-sol and gpt-6-luna, and SiliconANGLE confirms both were live for developers on day one.
What the two models are for

Sol is the mid-tier model and Luna is the small, cheap one, and OpenAI says GPT-6 Astra “continues to be our best model across the board.” Sol and Luna sit underneath Astra in the lineup.
The claim OpenAI leads with is accuracy. On its own internal evaluation, Sol “makes about half as many mistakes as its predecessor,” and that’s OpenAI grading OpenAI. The style notes are more concrete: “more clarity, less jargon, fewer odd turns of phrase, fewer low-value details, and slightly shorter answers overall.” OpenAI also published benchmark comparisons against Claude, and those are its own numbers.
Anthropic released Claude Opus 5.5 the same day, which is the same pattern I wrote about when GPT-6 Astra and Fable 5 shipped a week apart.
The price cut

API pricing is half what GPT-5.6 cost. Sol is $2 per million input tokens and $10 per million output, down from $4 and $20 for GPT-5.6 Sol, and Luna is $0.10 and $0.50, down from $0.20 and $1.20. Those figures come from OpenAI’s announcement, which calls it 50% cheaper than the GPT-5.6 promotional pricing.
If you don’t pay per token, this reaches you as Luna landing in the desktop app for Free and Go users, because cheap models are what make a free tier possible. OpenAI hasn’t published any change to ChatGPT message limits.
Where I land on it
I haven’t touched Sol or Luna yet, but I’ve used GPT-5.6 Sol a lot and it was really solid. I’m a fan of the Sol and Astra models, and I think they give Opus and Fable a run for their money.
That’s also why I’m moving off Claude Max 20x to Max 5x plus a ChatGPT Pro plan for Codex. The two companies compete closely enough now that we’re the beneficiaries, and there’s no reason to be locked into one brand when several of them do the same job. I still feel ChatGPT and Codex give more bang for the buck, especially after the Claude Code limit cut. If you built anything on custom GPTs, that’s changing too, so check that before you plan around Sol.
Frequently asked questions
Can I use GPT-6 Sol in regular ChatGPT?
Not yet. OpenAI says the models are in ChatGPT Work and Codex, and Luna is in the desktop app for free users. Regular chat still runs the previous lineup.
Is GPT-6 Sol better than GPT-6 Astra?
No. OpenAI says Astra is still its best model. Sol is the mid-tier option and Luna is the small, cheap one.
Do I need a paid plan to try GPT-6 Luna?
No. Free and Go users get Luna in the desktop app. Sol needs Plus, Pro, Business, Enterprise, or Edu.
Related reading: GPT-6 Astra got worse after launch | Claude Code usage limits are dropping 17% | OpenAI is retiring custom GPTs | New to AI? Start here
WHO WROTE THIS
Moses Smith. I write Everyday AI for people who aren’t engineers. I go try the tools, then tell you honestly whether they were worth it. Sometimes the answer is no, and that’s kind of the point.
This blog is free and has no ads. If it saved you some time, you can buy me a coffee.









Leave a Reply