GPT-6 Sol and Luna launch in Work, Codex, and the API

Quick answer

OpenAI announced GPT-6 Sol and GPT-6 Luna on September 22, 2026, with lower API token rates than their GPT-5.6 counterparts. They are rolling out in ChatGPT Work and Codex, but not regular Chat. Sol targets complex work; Luna targets narrower, repeatable tasks. This source-based launch guide explains access and costs, not independently measured performance. Neither exact model is verified in Chat AI's catalog.

Download Chat AI Opens the official App Store or Google Play for your device.

Where are GPT-6 Sol and Luna available?

OpenAI lists a rollout to Plus, Pro, Business, Enterprise, and Edu users in Work and Codex. Free and Go users get access to Luna in the desktop app, subject to rollout. Check the model picker in the product you actually use: the announcement does not put either model in regular Chat.

Sources: OpenAI, OpenAI

Sol or Luna: start with the task, not a ranking

OpenAI positions Sol for demanding coding and work involving multiple steps and tools. It positions Luna for well-defined jobs such as extracting fields, organizing information, or producing structured summaries. These are provider recommendations, not proof that either model will outperform your current choice on a particular document or project.

Sources: OpenAI, OpenAI

API prices are lower, with separate cache and long-input rates

The following standard US-dollar prices are per million tokens for requests with no more than 272,000 input tokens. Input is the material sent to the model; output is what it generates. Cached input reuses previously processed material.

GPT-6 Sol: $2 input, $0.20 cached input, and $10 output. Cache writes cost $2.50.
GPT-6 Luna: $0.10 input, $0.01 cached input, and $0.50 output. Cache writes cost $0.125.
For comparison, GPT-5.6 Sol is listed at $4 input and $20 output; GPT-5.6 Luna at $0.20 input and $1.20 output.

Sources: OpenAI, OpenAI, OpenAI

Do lower rates mean more included messages?

OpenAI publishes estimated local-message ranges for ChatGPT plans, not a fixed entitlement for every task. Model choice, context, reasoning, tools, retrieval, and caching all affect allowance consumption. Local and cloud tasks share the plan allowance, and weekly limits may apply. Check your usage dashboard for the current balance and reset time.

Sources: OpenAI

Model specifications do not determine an app's upload limits

Both API model pages list text and image input, text output, a 1,050,000-token context window, up to 922,000 input tokens, and up to 128,000 output tokens. Those are API specifications, not a promise that an app accepts a file of any size or exposes every supported tool.

Sources: OpenAI, OpenAI

GPT-6 Sol and Luna are not verified in Chat AI

Chat AI's published catalog lists GPT-5.6 Sol and GPT-5.6 Luna, not these GPT-6 versions. Do not download Chat AI expecting either new model. If you want a multi-model app for everyday work, check exact model names, document tools, and plan limits before choosing one.

Sources: Chat AI

Frequently asked questions

What readers usually ask

Why can't I see GPT-6 Sol or Luna in regular ChatGPT Chat?

This release is for Work and Codex, not regular Chat. Access also depends on rollout, your plan, client, and workspace controls.

Can I try GPT-6 Luna without a paid plan?

OpenAI lists desktop-app access for Free users, subject to rollout. That does not make API calls free or establish availability in other apps.

Are the GPT-5.6 models being removed immediately?

No. OpenAI's model guide says GPT-5.6 Sol, Terra, and Luna remain available during the rollout. Check the exact version in the picker when comparing results.

Evidence