GPT Luna — the fast GPT-5.6 entry tier | Lexicon
Current
“Luna” is the name of a line. The current flagship is GPT Luna version 5.6 (released July 9, 2026); on July 30, 2026 OpenAI additionally cut the line’s API prices by 80 percent. Details live in the glossary entry GPT Luna. For real-time voice there is also GPT-Live-1 — that is not a text model in the Sol/Terra/Luna tiering.
What GPT Luna is
Luna is the bottom of the three tiers in OpenAI’s GPT-5.6 family: the entry tier, built for speed and low cost per request. Above it sits Terra as the balanced middle, at the top Sol for the hardest work. Luna gives up reasoning and coding depth in exchange — which you do not need for simple, high-volume tasks anyway.
The name stays stable across generations while the version number moves up. So “Luna” means the permanent entry slot, not a single model.
Positioning
Luna sits in the volume band: the family’s highest speed and cost efficiency, in exchange for less depth on reasoning and coding. The suitability profile shows it clearly — a peak on speed and cost, more reserved in the demanding disciplines.
Eignung 0–5 · redaktionelle Einordnung, kein Benchmark
The axes are an editorial assessment, not a benchmark. They help with rough orientation but do not replace your own test on the concrete use case.
Use profile
Luna is the tier for everything that has to run at scale and fast:
- Classification and routing — sorting tickets, sending requests down the right path.
- Summaries of long texts in series.
- Simple chat assistants where response time and cost are what count.
- High-volume batch pipelines, where throughput per euro is the decisive figure.
As soon as a task demands multi-step reasoning or reliable coding across long chains, Terra is the better pick — and for the hardest cases, Sol.
For real-time voice the GPT-5.6 tiering is not the answer: for that there is GPT-Live-1, a dedicated voice model that hands more complex requests to a stronger text model in the background.
Selection guide
The positioning map places Luna against its family siblings and one external fast model:
Redaktionelle Einordnung, kein Benchmark
Rule of thumb: start with Luna. If the tier handles a task reliably, you do not need anything pricier. Only when quality or reasoning depth falls short do you move up to Terra, and to Sol for the top end. Because the market turns over fast, model choice belongs in one central, configurable place in your system.
More in the glossary
- GPT Luna — current version, price, context window and benchmarks
- GPT Terra — the balanced middle of the family
- GPT Sol — the frontier tier for the hardest tasks
FAQ
- Luna is the entry tier of the GPT-5.6 family: faster and cheaper than Terra and Sol, but with less reasoning and coding depth. Terra is the balanced middle, Sol the flagship for the hardest tasks.
- GPT Luna, officially GPT-5.6 Luna, released July 9, 2026 alongside Terra and Sol. On July 30, 2026 OpenAI additionally cut the line's API prices by 80 percent.
- No. GPT-Live-1 is OpenAI's real-time voice model for ChatGPT, not a text model in the Sol/Terra/Luna tiering. It hands more complex requests to a stronger text model in the background.
- Classification, routing, summaries, simple chat assistants and high-volume batch pipelines — anywhere throughput and cost per request come before maximum reasoning depth.
- No. Like the entire GPT-5.6 family, Luna is available only as a hosted service via the OpenAI API and cloud platforms such as Amazon Bedrock.