GPT-6 explained: Astra, Sol, Terra and Luna

GPT-6 is OpenAI’s current family of models, released from September 2026. It has four tiers: Astra at the top, then Sol, Terra and Luna. This guide explains what each GPT-6 model is for, when it arrived, what it costs and which to use.

What each GPT-6 model is for

Each tier has a different job:

• GPT-6 Astra is the top model. OpenAI calls Astra its “most intelligent and aligned model yet”, with state-of-the-art capabilities across computer use, coding, cybersecurity and science.
• GPT-6 Sol and Luna bring frontier intelligence to everyday work, each with a different balance of capability and cost. OpenAI’s Decisions API, built for “fast, focused, low-cost decisions”, runs on GPT-6 Luna.
• GPT-6.1 Sol offers near-Astra intelligence for coding, computer use and professional work.
• GPT-6.1 Terra brings fast, capable computer use to everyone. Computer use means the model operates websites and software on screen.

The Sol, Terra and Luna names first appeared with GPT-5.6 in July. The model names gpt-5.6-sol, gpt-5.6-terra and gpt-5.6-luna date from 9 July 2026. Astra is new with GPT-6.

For help choosing between models and tuning reasoning effort, read OpenAI’s model guide for the GPT-6 family.

When each model was released

Model	Announced	Model name
GPT-6 Astra	3 September 2026	gpt-6-astra
GPT-6 Sol	22 September 2026	gpt-6-sol
GPT-6 Luna	22 September 2026	gpt-6-luna
GPT-6.1 Sol	29 September 2026	gpt-6.1-sol
GPT-6.1 Terra	26 October 2026	gpt-6.1-terra

Dates are from OpenAI’s announcements. Model names and versions are as listed in Microsoft’s Azure OpenAI documentation.

How much GPT-6 costs in the API

GPT-6.1 Sol costs one-fifth of Astra’s standard API input and output token prices.

The table below shows Microsoft Foundry’s Global Standard prices for the same models. Prices are in US dollars per million tokens, for short prompts.

Model	Input	Cached input	Output
GPT-6 Astra	US$10	US$1	US$50
GPT-6 Sol	US$2	US$0.20	US$10
GPT-6.1 Sol	US$2	US$0.10	US$10

Long prompts cost more. On Microsoft Foundry, long-context input is US$20 per million tokens for Astra and US$4 for Sol. Check OpenAI’s own pricing page before you budget for Terra or Luna, and for any change since this was written.

GPT-6 also brought better prompt caching, with explicit cache breakpoints. Caching cuts the cost of sending the same instructions or documents many times.

Where GPT-6 Astra stands on safety

GPT-6 Astra is OpenAI’s most capable broadly deployed model. It is also the first OpenAI model to reach the “Critical” level of cybersecurity capability under the company’s Preparedness Framework.

What changed in ChatGPT

• GPT-6 in everyday chats. On 7 October, GPT-6 began rolling out in ChatGPT globally. It came with Intelligent UI, which adds visuals and interactive elements to answers.
• ChatGPT Work. Since 9 July, ChatGPT Work has offered an agent that takes action across your apps and files. It can stay with a project for hours.
• Ads in Southeast Asia. On 23 September, ChatGPT Ads expanded to Southeast Asia and Taiwan, reaching more than 60 countries in total.

Plans and usage limits change often. Check OpenAI’s pricing page for which models your ChatGPT plan includes.

Which GPT-6 model to use

• Luna for sorting, tagging, extraction and short replies at high volume.
• Terra for agents that operate websites and software.
• Sol for coding, research and multi-step analysis.
• Astra for the hardest problems, where accuracy matters more than cost.

Start with the cheapest model that does the job well. Move up only for the steps that need it. Test each choice on 20 or so real examples from your own work before you commit.

For a comparison with Claude and Gemini, read Claude, GPT-6 or Gemini 4? Choosing an AI model. For the month’s wider news, read October 2026 AI updates.