Back to glossary

Term

GPT Luna

GPT Luna (GPT-5.6 Luna) is the smallest, fastest and most affordable model of OpenAIs GPT-5.6 family, released in July 2026. It is built for cost-sensitive, high-volume workloads and offers a context window of roughly 1.05 million tokens.

GPT Luna — explained in detail

GPT Luna is the official model name for GPT-5.6 Luna, the smallest member of OpenAIs GPT-5.6 family. This family was released on 9 July 2026 and comprises three tiers, ordered from cheapest to most capable: Luna, Terra and Sol. Luna is the fastest and most affordable variant and roughly corresponds to the nano tier of earlier GPT-5 generations.

Key data according to the provider: a context window of around 1,050,000 tokens (of which up to 922,000 input tokens), a maximum output length of 128,000 tokens and a knowledge cutoff of 16 February 2026. Access is proprietary via the OpenAI API and platforms such as Amazon Bedrock. The model supports streaming, structured outputs, function calling, image input, web search and prompt caching, among other features.

Luna is deliberately tuned for cost efficiency. Its API pricing sits at the low end (around 0.20 US dollars per million input and 1.20 US dollars per million output tokens); on 30 July 2026, OpenAI cut the price by a further 80 percent. This makes Luna specifically aimed at high-volume applications.

Example / practical relevance

GPT Luna is suited to tasks where large volumes of requests must be processed cheaply — such as classification, summarisation, simple chat assistants or the pre-processing of large amounts of text. Its large context window also allows extensive documents to be handled in a single pass without switching to a more expensive model.

Distinction

Within the GPT-5.6 family, Luna is the entry tier: faster and cheaper than the mid-tier Terra and the flagship Sol, but less capable on the most demanding reasoning tasks. Against models from other providers (such as Claude or Gemini), Luna positions itself primarily on price and throughput rather than on maximum capability.

See everything in one place:GPT Luna