GPT-5.6 Luna
OpenAI · released Jul 9, 2026
- Type
- Closed
- Input
- $1.00
- Output
- $6.00
- Cached
- None held
List price · per 1M tokens · OpenAI at 1.1M context · machine-readable source ↗ · a reseller below undercuts it; the table carries the spread
Our take
Written Sep 17, 2026GPT-5.6 Luna is a hosted-only model from OpenAI: we list no download for it, so using it means choosing a host. It scores strongly on set-piece mathematics, reasoning and coding, and its listed offers vary widely in price, so the host you pick matters as much as the model.
Use it for self-contained coding, mathematics and reasoning tasks, or for long documents that would otherwise need splitting. Compare the listed offers before you commit, because the same model is sold at rates that differ by a factor of twenty. Skip it if you need to run the model on your own hardware, or if you need fixes landed in an existing codebase without review.
The case for it
- 87.2% on LiveBench mathematics and 85.64% on reasoning, both averages over competition-style and olympiad problems rather than applied work.
- 82.92% on LiveBench coding, which covers code generation and completion rather than fixing issues in an existing project.
- The request capacity holds a long report or a stack of documents beside the question, though reliable recall across all of it is unverified in our data.
- The cheapest listed offer is a fraction of the dearest, so the host you pick changes the bill by a factor of twenty.
The case against it
- 48.43% on LiveBench agentic coding, measured inside an agent harness, against 82.92% on LiveBench coding, so driving tools and recovering from failures is a different proposition from writing code to a specification.
- 60.12% on LiveBench instruction following, which covers constrained-rewriting tasks, against 87.2% on mathematics, so tightly specified output formats are where it is most likely to need checking.
- We list no download for it, so using it means choosing a host and accepting that host's terms and availability.
How good is it?
A text model from OpenAI built for calling tools to carry out requests.
- calling tools to carry out requestsArena Agent · Tool use · 9th of 55
EverydayGeneral questions and everyday reasoning
Arena Text (overall)45th of 168 · 1454
CodingWriting and fixing code on its own
Arena Coding51st of 168 · 1501
AgenticPlanning, calling tools, staying on task
Arena Agent30th of 55 · −0.008
WritingDrafting and rewriting prose
Arena Creative Writing54th of 168 · 1412
Placings on Arena's agent boards, from live sessions people ran themselves. A model can lead on one of these and sit mid-field on the others.
Boards this model appears on that none of the ratings above are built on.
Every published score for this model20 scoresEvery figure we hold, from 20 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 4 hours and 2 months ago — each listing carries its own date.
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
- $0.20 in / $1.20 out
- Context served
- 1.1M
- Throughput
- Not measured
| Provider | In / out per 1M tokens | Context | Throughput | Trains on prompts | Logs prompts | Zero retention |
|---|---|---|---|---|---|---|
| OpenAIflex tierThrough OpenRouter | $0.10 / $0.60checked 4 hours ago | 1.1M128K max reply | 48 tok/s | No | Yesunknown period | Unknown |
| OpenRouterOpenRouter's own listing | $0.20 / $1.20checked 4 hours ago | 1.1M | not measured | Unknown | Unknown | Unknown |
| Microsoft Azure AIeuThrough OpenRouter | $0.22 / $1.32checked 4 hours ago | 1.1M128K max reply | 66 tok/s | No | No | Confirmed |
| Amazon Bedrockus-east-1Through OpenRouter | $0.22 / $1.32checked 4 hours ago | 1.1M128K max reply | 84 tok/s | No | No | Unknown |
| Microsoft Azure AIusThrough OpenRouter | $0.22 / $1.32checked 4 hours ago | 1.1M128K max reply | 26 tok/s | No | No | Confirmed |
| OpenAIfast tierThrough OpenRouter | $0.40 / $2.40checked 4 hours ago | 1.1M128K max reply | 38 tok/s | No | Yesunknown period | Unknown |
| OpenAIDirect | $1.00 / $6.00checked 2 months ago | 1.1M | not measured | No | Yesunknown period | Unknown |
| Microsoft Azure AIThrough OpenRouter | $1.00 / $6.00checked 52 days ago | 1.1M | not measured | No | No | Confirmed |
Microsoft Azure AI
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 | ✓ | ✓ | ✓ |
| Microsoft Azure AIeuThrough OpenRouter | ✓ | ✓ | ✓ |
| Amazon Bedrockus-east-1Through OpenRouter | ✓ | ✗ | ✗ |
| Microsoft Azure AIusThrough OpenRouter | ✓ | ✓ | ✓ |
| OpenAIfastThrough OpenRouter | ✓ | ✓ | ✓ |
| OpenAIDirect | ✓ | ✓ | ✓ |
| Microsoft Azure AIThrough OpenRouter | ✓ | ✓ | ✓ |
Tool calling: 8 of 8 listings say yes. JSON output: 7 of 8 listings say yes, 1 says no. Strict schema: 7 of 8 listings say yes, 1 says no.
Models people weigh against GPT-5.6 Luna
When we formed this view
Recent changes
Each 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-5-6-luna