OpenAI introduced GPT-6 Sol and GPT-6 Luna on September 22. They sit below GPT-6 Astra in its model family. Think of choosing a vehicle for a trip: Sol is the stronger option for difficult work, while Luna is the lighter, cheaper choice when you have many simpler jobs. The right one depends on the task you give it.
What changed?
OpenAI says both models bring improvements from its GPT-6 work to lower-cost tiers. It reports better results than their GPT-5.6 predecessors on several coding, computer-use, and business workflow tests. The company lists API prices of $2 per million input tokens and $10 per million output tokens for Sol, and $0.10 and $0.50 for Luna. Those are usage prices for developers; a long task with many steps can still consume substantial tokens.
Where can you use them?
OpenAI says Sol and Luna are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users, and in its API as gpt-6-sol and gpt-6-luna. Free and Go users can access Luna in the desktop app. At launch, the models are not yet in standard Chat, and rollout through the day may delay when you see them. Having a subscription does not make every named model available in every product view.
How should you judge the claims?
Try the same task with each model: for example, ask one to explain a small code change, then inspect the code and run tests. Compare the result and the effort needed to fix it. OpenAI's published scores use specific tools and effort settings; competitor scores come from public reports, and its internal factuality test is drawn from conversations where people flagged errors. Those tests are useful clues, not a guarantee that Sol or Luna will be better on your own work.