Updated

A smooth pale moonstone crescent balanced upright on its edge on a dark background, abstract hero artwork for GPT-6 Luna

OpenAI

GPT-6 Luna

OpenAI's most efficient GPT-6 model for focused, high-volume work, and Super Agent's default.

Chat with GPT-6 Luna in Super AgentView pricing

What is GPT-6 Luna?

GPT-6 Luna is the most efficient model in OpenAI's GPT-6 series, released on September 22, 2026 and positioned below GPT-6 Sol.

OpenAI built it for focused, repeated work with a clear goal, such as extracting invoice fields, classifying requests or writing structured summaries. It can also think hard when asked: OpenAI reports 66.6% on DeepSWE v1.1 at max effort and says that at higher effort it matches GPT-5.6 Sol on factuality. It is the default model in Super Agent, with reasoning from off to max.

Provider
OpenAI
Context window
1.05M tokens
Input
Text, Image
Reasoning
Off to max
Max output
128K tokens
Knowledge cutoff
May 18, 2026
Released
Sep 22, 2026

How GPT-6 Luna performs

Published by OpenAI

Results OpenAI reported for GPT-6 Luna. Each lab runs its own setup, so compare these with other models only inside the same table.

Benchmark results OpenAI published for GPT-6 Luna
BenchmarkScoreSetting
DeepSWE66.59%max effort
Agents' Last Exam50.89%max effort
OSWorld 2.052.68%offline set, max effort
FrontierCode42.42%max effort
AutomationBench20.7%max effort
Factual error rate on difficult prompts (lower is better)7.56%max effort

Source: openai.com

What you can do with GPT-6 Luna

  • The default model in Flashloop's Super Agent

  • Reasoning from off to max, so it can reply instantly or think hard

  • Matches GPT-5.6 Sol on factuality at higher effort, according to OpenAI

  • 5.4 points above GPT-5.6 Luna on AutomationBench at high effort

  • Takes images in a 1.05M-token context window

How to prompt GPT-6 Luna

  • Turn reasoning off for instant replies when you only need ideas or rewrites.
  • Raise effort to high before asking it to plan a multi-step project.
  • Keep prompts short and direct; Luna is built for quick, focused turns.

Where GPT-6 Luna fits best

  • Fast back-and-forth chat

  • Quick prompt ideas for images and videos

  • Sorting, tagging and summarizing many items

  • Light agent tasks where speed matters

How GPT-6 Luna compares

GPT-6 Luna compared with other chat models on Flashloop
ModelLMArena textLMArena codingContext windowReasoning
GPT-6 Luna1442#911507#651.05M tokensOff to max
Grok 4.71441#961493#91500K tokensLow to max
DeepSeek V4 Flash1436#1051484#1061M tokensOff, medium, high, extra high
GPT 5.6 Luna1453#751502#761.05M tokensOff to max
Gemini 3.5 Flash Lite1455#711502#771M tokensLow, medium, high

Good to know before you start

  • OpenAI positions it below GPT-6 Sol and GPT-6 Astra on demanding work
  • Replies in text; any image or video you ask for is rendered by one of Flashloop's generation models
  • OpenAI's system card notes it may refuse more requests than other GPT-6 models, including legitimate ones

Start chatting with GPT-6 Luna

Chat with GPT-6 Luna in Super Agent

  1. Select GPT-6 Luna

    Open Super Agent on Flashloop and choose GPT-6 Luna in the model picker.

  2. Enter your prompt

    Ask for what you need: an answer, a plan, or the images and videos you want the agent to make.

  3. Send and keep going

    Send the message. GPT-6 Luna answers in Super Agent and runs any generations you asked for.

Questions about GPT-6 Luna

Is GPT-6 Luna the default model in Super Agent?

Yes. Super Agent uses GPT-6 Luna unless you pick another model in the model picker. It starts at low reasoning effort.

How does GPT-6 Luna compare with GPT-6 Sol?

Luna is the faster, cheaper tier below Sol. OpenAI aims Luna at focused, high-volume tasks with a clear goal and Sol at complex coding and agent work.

Can GPT-6 Luna reply without thinking?

Yes. You can set reasoning to off for instant replies, or go up to max for hard tasks.

What is GPT-6 Luna Pro?

The same model run in OpenAI's pro reasoning mode. It does more work on each request, so it is slower and better kept for hard questions.

More chat models on Flashloop

Browse all models