GPT-6 Luna
OpenAI · released Sep 22, 2026
- Type
- Closed
- Input
- $0.10
- Output
- $0.50
- Cached
- $0.010
List price · per 1M tokens · OpenAI at 922K context · machine-readable source ↗
Our take
Written Sep 29, 2026GPT-6 Luna is a hosted-only model released in September 2026, with no download on offer and no disclosed licence, so using it means picking a host. Its measured results are mid-field overall but noticeably better on coding and web-app building, and weak on instruction following.
Reach for it on coding and web-app building, where its Arena Code (WebDev) placing is the strongest signal in front of us, and for long documents that would otherwise need splitting up first. Eight hosted offers mean you can move the workload if one host's terms or rates stop suiting you. Skip it if you need to run the model yourself, or if instruction following is central to your task.
The case for it
- 20th of 95 on Arena Code (WebDev) via Max as of 25 Sep 2026, and 31st of 168 on Arena Coding via Max as of 25 Sep 2026 — both human-preference boards, so they record which answer people preferred rather than whether the code ran.
- 89.12% on LiveBench Mathematics, which covers competition and olympiad maths tasks rather than applied arithmetic.
- The request capacity takes long documents in one go, so files need not be split up first; reliable recall across all of it is unverified in our data.
- Eight hosted offers across four companies, so a rate rise or a terms change at one is a reason to move the workload rather than a bill to absorb.
The case against it
- 56th of 58 on LiveBench Instruction Following via Max as of 25 Jun 2026, a board of constrained-rewriting tasks, so anything that depends on following a tight format is a risk.
- 55th of 168 on Arena Text (overall) via Max as of 25 Sep 2026 — mid-field on a board that records which answer people preferred rather than whether it was correct.
- We list no download for it and the licence is not disclosed, so using it means choosing a host and checking terms with them.
How good is it?
A closed text model from OpenAI built for writing and completing code.
- writing and completing codeArena Coding · 31st of 168
EverydayGeneral questions and everyday reasoning
Arena Text (overall)55th of 168 · 1442
CodingWriting and fixing code on its own
Arena Coding31st of 168 · 1515
AgenticPlanning, calling tools, staying on task
Not yet scored on Arena Agent. It is on LiveBench Agentic Coding, in 33rd of 58 with 51.21.
WritingDrafting and rewriting prose
Arena Creative Writing65th of 168 · 1403
Boards this model appears on that none of the ratings above are built on.
Every published score for this model15 scoresEvery figure we hold, from 15 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Where to rent it
Prices checked between 55 min and 58 min ago — each listing carries its own date.
Microsoft Azure AI, through OpenRouter
Why this differs from the header. The strip above quotes OpenAI's own list price; this is the cheapest live offer, whoever is serving it — a reseller undercutting a lab is ordinary commerce, not an error.
- per 1M tokens
- $0.10 in / $0.50 out
- Context served
- 1.1M
- Throughput
- ~54 tok/s
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenAIflex tierThrough OpenRouter | $0.050 / $0.25checked 56 min ago | 1.1M128K max reply | 35 tok/s | No | Yesunknown period | Unknown |
| OpenRouterOpenRouter's own listing | $0.10 / $0.50checked 58 min ago | 1.1M | not measured | Unknown | Unknown | Unknown |
| OpenAIDirect | $0.10 / $0.50checked 55 min ago | 922K128K max reply | 66 tok/s | No | Yesunknown period | Unknown |
| Microsoft Azure AIThrough OpenRouter | $0.10 / $0.50checked 56 min ago | 1.1M128K max reply | 54 tok/s | No | No | Confirmed |
| Amazon Bedrockus-east-1Through OpenRouter | $0.11 / $0.55checked 56 min ago | 1.1M128K max reply | 86 tok/s | No | No | Unknown |
| Microsoft Azure AIusThrough OpenRouter | $0.11 / $0.55checked 56 min ago | 1.1M128K max reply | 26 tok/s | No | No | Confirmed |
| Microsoft Azure AIeuThrough OpenRouter | $0.11 / $0.55checked 56 min ago | 1.1M128K max reply | 83 tok/s | No | No | Confirmed |
| OpenAIfast tierThrough OpenRouter | $0.20 / $1.00checked 56 min ago | 1.1M128K max reply | 94 tok/s | No | Yesunknown period | Unknown |
OpenRouter
Amazon BedrockAcross the 8 listings we hold: 7 say they do not train on prompts, 0 say they do and 1 does not say. 3 appear in the zero-retention registry we check; the rest are unknown to us.
What each host's API supports
From the parameter list each endpoint publishes. Streaming is omitted: nothing we hold reports it, for any model.
| Provider | Tool calling | JSON output | Strict schema |
|---|---|---|---|
| OpenAIflexThrough OpenRouter | ✓ | ✓ | ✓ |
| OpenRouterOpenRouter's own listing | ✓ | ✓ | ✓ |
| OpenAIDirect | ✓ | ✓ | ✓ |
| Microsoft Azure AIThrough OpenRouter | ✓ | ✓ | ✓ |
| Amazon Bedrockus-east-1Through OpenRouter | ✓ | ✓ | ✓ |
| Microsoft Azure AIusThrough OpenRouter | ✓ | ✓ | ✓ |
| Microsoft Azure AIeuThrough OpenRouter | ✓ | ✓ | ✓ |
| OpenAIfastThrough OpenRouter | ✓ | ✓ | ✓ |
Tool calling: 8 of 8 listings say yes. JSON output: 8 of 8 listings say yes. Strict schema: 8 of 8 listings say yes.
Models people weigh against GPT-6 Luna
When we formed this view
Recent changes
What moved
first indexed by our pipelineEach date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
- 1 of 8 listings does not say whether it trains on prompts.
- We hold no batch or off-peak rate for any of its listings.
Licence and identifiers
What the licence allowsWe hold no licence record for this model. Inside are the identifiers you need to pull it — its Hugging Face repo where we have one, our slug and a machine-readable card.
Licence
We hold no licence record for this model, and no record of published weights either — so we can neither summarise its terms nor point you at the weights.
Identifiers
- Takes in, gives back
- Text, images and documents in, text out
- Catalogue slug
- openai-gpt-6-luna