Identity and current terms
| Provider / API ID | OpenAI / gpt-5.6-luna |
|---|---|
| Context / maximum output | 1,050,000 / 128,000 tokens |
| Inputs → output | Text and images → text |
| Reasoning effort | None, low, medium, high, xhigh, max; medium is documented as default |
| Standard short-context price | $0.20 uncached input, $0.02 cached input, $1.20 output per 1M tokens |
| Long-context condition | Requests above 272K input tokens use 2× input and 1.5× output rates for the full request |
Checked against the official model and pricing pages on 12 September 2026. Record the requested and returned model IDs; the inspected listing does not establish a separate dated snapshot.
Why start here?
The low token price makes a controlled first run inexpensive. For an illustrative 80K uncached input plus 10K billed output tokens, the short-context token cost is $0.028. That is not a forecast of the tokens an agent will use and says nothing about whether its patch will be accepted.
Luna supports function calling, structured outputs, streaming, and Responses API tools relevant to coding agents. Those features establish integration possibilities—not repository correctness.
Where the evidence stops
Provider-reported coding evaluations make Luna credible enough to test, but they do not provide a matched AIModel.Reviews dataset of accepted patches, retries, total cost, and human review. A million-token context window does not prove reliable use of a million tokens of repository material.
Evaluation checklist
- Pin the starting commit and write acceptance checks before the run.
- Record reasoning effort, tools, permissions, and all billed usage.
- Keep failed attempts and human interventions.
- Compare Luna at a higher effort before changing both model and settings at once.
- Escalate only the task category that exhibits a repeatable failure.
Compare the premium in Luna versus Terra or start from the repository-coding buyer guide.
Reader questions
Frequently asked questions
How much does GPT-5.6 Luna cost?
The direct OpenAI standard short-context rates checked on 12 September 2026 are $0.20 per million uncached input tokens, $0.02 cached input, and $1.20 output. Other conditions can change the bill.
Is GPT-5.6 Luna proven enough for repository coding?
No universal adequacy claim is supported. It is a low-cost candidate to test against a repository-specific acceptance bar.
Primary reading
Sources and update conditions
Rates, availability, model identity, and provider behavior can change. Recheck the linked sources before making a consequential purchase or sending repository data.