Automating tasks with top-tier AI gets expensive once you are running thousands of messages, emails or documents a month. Sol and Luna are the cheaper, faster versions of GPT-6. OpenAI says it trained them with the same methods as Astra, its top model, and that their API costs half of what GPT-5.6's did.
Who this is NOT for
Not for you if your data cannot leave your office: this runs on OpenAI's cloud. Look at Ollama for that. If you have nobody to wire up an API, the price table does not apply to you. And if this is the task that matters most this quarter, OpenAI itself says that is a job for Astra, not Sol.
Read the fine print. The '50% cheaper' is measured against GPT-5.6's promotional pricing, not its regular rate. And every comparison against Claude Opus 5 and Fable 5.1 comes from benchmarks OpenAI publishes itself. For a small business the most useful number is about Luna: OpenAI says that on factuality, at higher effort, it matches GPT-5.6 Sol for roughly a hundredth of the cost. The usual mistake is starting with the big model out of habit. Test your repetitive task on Luna and move up to Sol only if Luna falls short.
Your first step
Open the announcement page and go straight to the 'Availability' section. It is the one we could not fully retrieve. Check there whether Sol and Luna are in your ChatGPT plan or API-only before you spend a dollar.
The GPT-6 Sol y GPT-6 Luna (OpenAI) implementation guide
The project belongs to openai.com and it is free: the link beside asks for nothing. The guide to set it up without getting lost is ours: step by step, common mistakes and a checklist. Leave your email and we will send it.
No spam. Unsubscribe anytime.