GPT-6 Luna

// Overview

GPT-6 Luna is the cheapest model in OpenAI's GPT-6 family, released on September 22, 2026 alongside the balanced GPT-6 Sol. OpenAI's API documentation calls it "Our most efficient model for focused, high-volume tasks", and GitHub's Copilot changelog describes it as "a lightweight, cost-efficient model for smaller, faster tasks and the lowest-cost option in the GPT-6 family". TechCrunch reports OpenAI aiming it at clerical work — summarizing documents, extracting information, and answering quick questions.

Luna accepts text and image input and returns text, with a 1,050,000-token context window and up to 128,000 output tokens — the same window as Astra and Sol. Its knowledge cutoff is May 18, 2026. It is served on the Chat Completions, Responses, and Batch endpoints and supports streaming, structured outputs, function calling, file search, image input, web search, and prompt caching.

OpenAI lists GPT-6 Luna at $0.10 per 1M input tokens, $0.01 per 1M cached input tokens, and $0.50 per 1M output tokens. That is half the GPT-5.6 Luna rate of $0.20 input and $1.20 output — a 58% cut on the output side. Long-context requests are billed at $0.20 input, $0.02 cached input, and $0.75 output per 1M tokens.

At launch Luna reached further into OpenAI's products than Sol did: alongside the ChatGPT API, Codex, and ChatGPT Work, TechCrunch reports it also going to the desktop app and to Free and Go users. GitHub added it to Copilot the same day on Pro, Pro+, Max, Business, and Enterprise plans — one tier lower than Sol, which starts at Pro+.

// Pricing

Down from the GPT-5.6 Luna rate of $0.20 input / $1.20 output. Long-context requests are billed at $0.20 input, $0.02 cached input and $0.75 output per 1M tokens. Regional data residency endpoints carry a 10% surcharge.

Pricing source ↗

// Strengths

• The lowest token price in the GPT-6 family at $0.10 input per 1M tokens
• 1.05M-token context window, matching the Astra and Sol tiers
• Prompt caching at $0.01 per 1M tokens, a tenth of the input rate
• Served on the Responses endpoint as well as Chat Completions and Batch
• Reaches Free and Go ChatGPT users, not just paid API accounts

// Best for

• Reach for it for high-volume clerical work — summarizing documents, extracting fields, answering short questions.
• Reach for it when a pipeline runs enough tokens that price, not peak capability, sets the budget.
• Reach for it in GitHub Copilot on a Pro plan, the lowest tier that carries a GPT-6 model.
• Reach for it for large batch jobs through the Batch endpoint, where the cached-input rate compounds.

// How to access

// GPT (Flagship / Thinking) — every version

The full lineage of the GPT (Flagship / Thinking) line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.