OpenAI now features three models: GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna. Astra is the most capable, Sol is nearly as good for less, and Luna is the cheapest for simple, high-volume jobs. If you are not sure, OpenAI says to start with Astra. The GPT-5.6 models that led the list until September 2026 are now the previous generation.1
The rest of OpenAI’s catalogue, over a hundred models, is a previous generation, a specialised variant, or something already scheduled for shutdown.2
- Astra, Sol, Luna. Those are the three models OpenAI features now, all from its GPT-6 generation.
- Astra is the ceiling — “our most capable model for the most demanding work”.
- Sol is the balance and Luna is the cheap, fast one.
- GPT-5.6 (Sol, Terra, Luna) is now previous generation. It is still sold and still works; it is just no longer what OpenAI puts first.
- Anything with
previewin the name can go in two weeks.
This page no longer lists model prices. The one price left (a voice session) was correct when this site last checked it on 22 Sep 2026. Prices change often and this site no longer updates them, so check the vendor’s own page before you rely on one: OpenAI pricing.
The three that matter
OpenAI released the three over four weeks. GPT-6 Astra came first, on 3 September 2026. GPT-6 Sol and GPT-6 Luna followed on 22 September. On 29 September OpenAI released GPT-6.1 Sol, a newer version of Sol, and the models page now features it in Sol’s place.34
| Model | What OpenAI says it is for | In plain words |
|---|---|---|
| GPT-6 Astra | “Our most capable model for the most demanding work.” | The strongest and most expensive. Use it when the job is hard and a wrong answer is costly. |
| GPT-6.1 Sol | “Near-Astra performance for complex work at a lower cost.” | The everyday workhorse for serious work: coding, long documents, agents (programs that carry out several steps on their own). |
| GPT-6 Luna | “Our most efficient model for focused, high-volume tasks.” | The cheap, fast one. Good for simple jobs you run thousands of times, like sorting or short summaries. |
OpenAI, Models, read at source 1 Oct 2026. For what each costs, see OpenAI’s pricing page (read 1 Oct 2026); this page does not list new prices.
The three are a ladder, not a menu. The question is which rung each part of your work needs, not which model is best. Start cheap, and move up only where the cheaper model gets it wrong.
A plain-English guide to the two newer models, including which ChatGPT plans got them, is on GPT-6 Sol and Luna, explained.
Where GPT-5.6 went
Until September 2026 the featured list was GPT-6 Astra plus three GPT-5.6 models: Sol, Terra and Luna. On 1 October 2026 the models page no longer features any of them. The only GPT-5.6 model it still shows is GPT-5.6 Cyber, a gated security model.1
They have not been switched off. OpenAI’s pricing page still lists gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna, and its deprecation page still names GPT-5.6 models as the replacements for older models it is retiring. No shutdown date for GPT-5.6 was listed on 1 October 2026.5
If your code already runs on GPT-5.6, nothing forces a move today. Test the GPT-6 model at the same rung before you switch, because a newer model can behave differently on the same prompt.
What they all do
OpenAI states one thing about the whole current family, which saves checking each model page: “All latest OpenAI models support text and image input, text output, multilingual capabilities, and vision.”
The GPT-6 models fit the same pattern: OpenAI says GPT-6 Sol and Luna “accept text and image inputs and generate text”.6
So the choice between the three is cost and capability depth, not feature availability. What is not uniform is price behaviour — long context roughly doubles every rate, and the detail is on what the OpenAI API actually costs.
How to read the other hundred
The full catalogue is mostly history and specialism. Four rough groups:
- Previous generations. GPT-5.6 Sol, Terra and Luna (featured until September 2026), GPT-5 through 5.5, GPT-4 and its variants, and the entire o-series. OpenAI marks these plainly — o3 is “succeeded by GPT-5”, and o4-mini is “succeeded by GPT-5 Mini”.
- Task-specific variants. Codex models for agentic coding, Realtime and Audio families for voice, Transcribe for speech-to-text, embeddings and moderation. The newest voice model, GPT-Live 1, became generally available on 10 September 2026 at $0.05 a minute for the voice session, with any backend model billed on top.7
- Gated specialist models (only approved organisations can use them). GPT-5.6 Cyber is described as “our most advanced cybersecurity model for authorized vulnerability research and security testing”, and GPT-Rosalind as “life sciences reasoning for approved organizations”. Both are access-controlled rather than open to any account. GPT-Rosalind became generally available to approved organisations on 8 September 2026, with billing from 5 October.8
- Open weights (the model files are published, so anyone can download and run them).
gpt-oss-120bis described as the “most powerful open-weight model, fits into an H100 GPU”, with a 20b sibling.
The word that should stop you
If a model name contains preview, do not build a product on it. OpenAI is direct about the risk: preview models “may be retired with much shorter notice, such as 2 weeks”, and it says it does not recommend them “for business-critical production workloads unless you can migrate on short notice”.9
Several models still listed in the catalogue already have shutdown dates attached. The notice periods, and what a retirement actually does to your calls, are on when AI models retire.
Against the others
Anthropic takes a similar approach with a shorter list: four current models (Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 4.5), the older ones listed separately as legacy, and a stated default.10
That comparison is on which Claude model to use, and the prices sit side by side on the four-vendor table.
What this page could not verify
- Any performance claim. Every description is OpenAI’s own, including “near-Astra performance” for Sol.
- Whether GPT-6 Sol (without the .1) stays available. The pricing page still listed
gpt-6-solon 1 Oct 2026, but the models page no longer features it, and no shutdown date was published. - Context window sizes. Held on individual model pages, not read here.
- Which model suits your task. That needs your evaluations, not a catalogue.
- Access requirements for the gated models. Described as approved or authorised without the process being stated.
Related on this site: GPT meaning: what Generative Pre-trained Transformer means
- 1OpenAI, Models, read at source 1 Oct 2026“If you’re not sure where to start, use GPT-6 Astra, our flagship model for complex reasoning and coding. Choose GPT-6.1 Sol to balance intelligence and cost, or GPT-6 Luna for cost-sensitive, high-volume workloads.” ↑afeatured models GPT-6 Astra, GPT-6.1 Sol and GPT-6 Luna; “GPT-5.6 Cyber” appears under specialised models. ↑b
- 2OpenAI, Models, read at source 16 Sep 2026. The page separates “Featured models” from the full catalogue.↑
- 3OpenAI, API changelog, read at source 1 Oct 202622 Sep 2026, “Released GPT-6 Sol and GPT-6 Luna. These reasoning models accept text and image inputs and generate text through the Responses and Chat Completions APIs.”; 29 Sep 2026, “Released GPT-6.1 Sol for complex coding and professional work at a lower cost than GPT-6 Astra.” ↑
- 4OpenAI, GPT-6 Astra system card, dated 3 Sep 2026, read at source 1 Oct 2026.↑
- 5OpenAI, Pricing and Deprecations, both read at source 1 Oct 2026GPT-5.6 models listed with prices; recommended replacements such as
gpt-5.6-terraandgpt-5.6-sol; no GPT-5.6 deprecation listed. ↑ - 6OpenAI, API changelog, 22 Sep 2026, read at source 1 Oct 2026.↑
- 7OpenAI, API changelog, 10 Sep 2026, read at source 22 Sep 2026“GPT-Live 1 is now generally available in the API.” and “Voice sessions cost $0.05 per minute, billed per second; backend model and tool usage is charged separately.” ↑
- 8OpenAI, API changelog, 8 Sep 2026, read at source 22 Sep 2026“GPT-Rosalind (
gpt-rosalind-research) is now generally available through the trusted-access program for approved internal life sciences research.” and “Billing begins on October 5, 2026.” ↑ - 9OpenAI, Deprecations, read at source 16 Sep 2026.↑
- 10Anthropic, Models overview, read at source 1 Oct 2026“If you’re unsure which model to use, start with Claude Opus 5.5 for most workloads.” ↑
OpenAI, Models, Pricing and Deprecations. All read at source on 16 September 2026.
OpenAI, Models, Pricing, Deprecations and API changelog, all read at source on 1 October 2026 · OpenAI, GPT-6 Astra system card, 3 September 2026 · Anthropic, Models overview, read at source on 1 October 2026.
Model lineups change monthly. Re-read before committing.
Free printable AI cheat sheets, and AI news every week.
Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.
SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK