What is GPT-6 Luna?
OpenAI released GPT-6 Luna on September 22, 2026, alongside GPT-6 Sol. It replaces GPT-5.6 Luna at roughly half the API price, and OpenAI positions it for efficient, repeatable work where speed and cost matter more than maximum depth: summarizing documents, extracting information, and answering quick questions.
One detail stands out: Luna has the newest knowledge cutoff in the GPT-6 family (May 18, 2026), newer than both GPT-6 Sol and the flagship GPT-6 Astra. For everyday questions about recent events, the cheapest model can know more than the expensive ones, although for anything current you should still turn on web search.
GPT-6 Luna specs and pricing
| Made by | OpenAI |
| Released | September 22, 2026 |
| API model ID | gpt-6-luna |
| Context window | 1.05M tokens (922K input) |
| Max output | 128,000 tokens |
| Knowledge cutoff | May 18, 2026 |
| Inputs | Text and images |
| API price | $0.10 input / $0.50 output per 1M tokens |
| In AimiChat | Fast quality · Free ($0) |
API prices are what developers pay OpenAI directly. In AimiChat you pay a flat plan price instead and do not need your own API key. Limits per message and per day depend on your plan.
What GPT-6 Luna is good at
- Everyday questions. Quick explanations, definitions, rewrites and short answers where you want a response in seconds rather than a long deliberation.
- Summaries and extraction. Pulling the key points, dates, names or figures out of a pasted document, email thread or article.
- Drafting. First drafts of emails, messages, outlines and social posts that you will edit yourself.
- Image questions. Reading a screenshot, a chart or a photo of a page and answering questions about it.
When to use a different model
- Hard multi-step reasoning. Proofs, tricky logic puzzles and decisions with many interacting constraints benefit from a larger model. Try Claude Sonnet 5 or GPT-6 Sol.
- Large coding tasks. Luna handles snippets and small fixes well, but multi-file changes and code review are a better fit for GPT-6 Sol or Claude Opus 5.5.
- Long, polished writing. For reports and long-form pieces where tone and structure matter, Claude Sonnet 5 usually needs fewer edits.
GPT-6 Luna benchmarks
| Benchmark | Score | Reported by |
|---|---|---|
| DeepSWE v1.1 (max effort) | 66.6% | OpenAI launch post, as reported by VentureBeat and Codersera |
These are the maker's own published results. Different companies run different test setups, so scores are only comparable within the same table from the same source.
Prompts to try with GPT-6 Luna
- Summarize this article in five bullet points, then list any claims that would need a source.
- Rewrite this email so it is shorter and friendlier, and keep every date and number exactly as written.
- Explain this error message in plain English and give me the most likely fix.
How to use GPT-6 Luna in AimiChat
- Open AimiChat and sign in with a free account.
- Choose the Fast quality in the model menu next to the message box.
- Send your message. You can switch quality in the same chat at any time, so you can start with a faster model and move up when a question needs more depth.
Compare GPT-6 Luna
- GPT-6 Sol vs GPT-6 Luna: when GPT-6 Sol is the better pick.
- Claude Sonnet 5 vs GPT-6 Luna: when Claude Sonnet 5 is the better pick.
- Claude Opus 5.5 vs GPT-6 Luna: when Claude Opus 5.5 is the better pick.
See every model in AimiChat side by side.
GPT-6 Luna FAQ
Is GPT-6 Luna free in AimiChat?
Yes. GPT-6 Luna powers the Fast quality, which is included on the Free plan with daily limits. Paid plans add higher limits and larger models.
What is the difference between GPT-6 Luna and GPT-5.6 Luna?
They are separate models. GPT-6 Luna is newer, has a later knowledge cutoff (May 2026) and costs about half as much through OpenAI's API ($0.10 / $0.50 per million tokens versus $0.20 / $1.20).
Is there a GPT-6 Terra?
No. OpenAI's GPT-6 family is Astra, Sol and Luna. Terra only exists in the older GPT-5.6 family.
How big is GPT-6 Luna's context window?
OpenAI lists 1.05 million tokens, with input capped at 922,000 tokens and up to 128,000 output tokens. The amount you can use in a single AimiChat message depends on your plan.
Sources
Specifications and prices come from the model makers and API listings below. Benchmark figures are reported by the company that made the model and are not independent tests.