GPT-5.6 Luna: Price, Context Window and When to Use It

GPT-5.6 Luna is the lightest model in OpenAI's GPT-5.6 generation, released on July 9, 2026 together with GPT-5.6 Sol and GPT-5.6 Terra. It keeps the family's 1.05 million-token context window and 128K-token maximum output, supports reasoning, vision, tools and structured output, and is priced for volume: about $0.20 per million input tokens and $1.20 per million output tokens on the standard API route. It is built for fast, high-volume jobs – classification, extraction, summaries, batch processing – rather than the hardest reasoning tasks.

Prices and specs below are as listed by API providers in October 2026 (OpenAI direct, Azure and AWS Bedrock routes, as aggregated by the Opper gateway). Check your provider's pricing page before you budget – OpenAI changes prices and tiers regularly.

GPT-5.6 Luna specs

	GPT-5.6 Luna
Released	July 9, 2026
Family	GPT-5.6 (Sol, Terra, Luna) – Luna is the lightest
Context window	~1.05 million tokens
Max output	128K tokens
Inputs	Text, images, PDF
Features	Reasoning, tool calling, structured output
Input price	$0.20 per 1M tokens (cache read $0.02)
Output price	$1.20 per 1M tokens
Where	OpenAI API, Azure (incl. EU / Sweden region), AWS Bedrock

On OpenAI's own route, flex processing (half price, slower) and priority processing (2x price) are also offered per request.

What GPT-5.6 Luna is good at

Luna is the "workhorse" tier. Its strengths are speed and price per call, which add up when you run millions of requests:

Classification and tagging – support tickets, product categories, moderation labels.
Extraction – pulling fields out of invoices, forms, PDFs or emails into JSON (structured output helps here).
Summaries – meeting notes, articles, chat logs.
Batch jobs – rewriting product descriptions, translating short texts, cleaning data.
The fast layer in a pipeline – Luna does the first pass or routing, and a bigger model handles only the hard cases.

Gateway descriptions also report that on agentic terminal (coding-agent) benchmarks Luna performs roughly on par with the earlier GPT-5.5 generation, which makes it usable for simple coding agents and scripts.

Where Luna falls short

The bigger GPT-5.6 tiers (Sol, Terra) pull ahead on recall over very long inputs and on deep, multi-step reasoning. Luna accepts a million tokens, but if your task is "find the one relevant clause in 800 pages", test a bigger tier before you trust Luna's answer. For hard reasoning, research or tricky code, use the larger models and keep Luna for the volume.

GPT-5.6 Luna vs Sol vs Terra

The three GPT-5.6 models share the same context window and feature set; they differ in depth, speed and price. A simple rule:

Luna – default for high-volume, short-context work where cost and latency matter.
Terra / Sol – step up when quality on long documents or multi-step reasoning matters more than price.

OpenAI has since released the GPT-6 generation (including GPT-6 Astra and GPT-6.1 Sol). If you are starting a new project, compare Luna with the newer models too – see our GPT-6.1 vs GPT-6 comparison. For many cheap bulk tasks, Luna remains a reasonable choice because of its price.

How to use GPT-5.6 Luna

API: call the model id gpt-5.6-luna through the OpenAI API (or your Azure / Bedrock deployment).
Keep prompts short and structured. Luna works best with clear instructions and a fixed output format – ask for JSON with named fields.
Use caching. Cached input is a tenth of the normal input price, so put the long, repeated part of the prompt (instructions, examples) first.
Test against a bigger tier on 50–100 real samples before switching a production job to Luna.

Writing a tight instruction is half the work with small models. Our free ChatGPT prompt generator turns a rough task description into a structured prompt you can adapt.

FAQ

When was GPT-5.6 Luna released? July 9, 2026, alongside GPT-5.6 Sol and GPT-5.6 Terra.

How much does GPT-5.6 Luna cost? About $0.20 per million input tokens and $1.20 per million output tokens on the standard route, with cached input at $0.02 per million. Azure's EU region lists slightly higher prices.

What is GPT-5.6 Luna's context window? About 1.05 million tokens, with up to 128K tokens of output.

Is GPT-5.6 Luna available in ChatGPT? It is mainly offered as an API model (OpenAI, Azure, AWS Bedrock). Which models appear in the ChatGPT app depends on your plan and changes often.

Is GPT-5.6 Luna good for coding? For simple scripts and agent loops, yes – its terminal-agent results are reported to be close to GPT-5.5. For large refactors or hard bugs, use a bigger model.

Is there a GPT-6 Luna? Yes, OpenAI's newer GPT-6 generation also lists a Luna model. GPT-5.6 Luna is the earlier one from July 2026.