What changed
OpenAI released GPT-6 Sol and GPT-6 Luna on September 22, cutting prices 50% from the old GPT-5.6 versions. Sol costs $2 per million input tokens and $10 output — down from $4 and $20. Luna costs $0.10 input and $0.50 output — down from $0.20 and $1.20. OpenAI confirmed the cuts are permanent, not a promotion. Both models are live in ChatGPT Work, Codex, and the API today. Sol handles complex tasks like coding. Luna handles high-volume repetitive work like summarizing or answering short questions.
What you can do with it this week
Pick the task your team does 100 times a month but hates doing. Luna is built for that: responding to common customer questions, pulling names and emails from forms, summarizing meeting notes, or tagging support tickets. Sol can handle the heavier lift: writing product descriptions from spec sheets, drafting proposals from your template and the client brief, or cleaning messy CSV files. Try one week of the repetitive task. Track how many you check vs. how many you just send. If you're checking every single one, the task isn't ready yet.
What to keep in your own hands
Anything with your customer's private data that Luna would need to store to answer. The approval to send a message that commits you to a price, a date, or a scope. The call on whether the AI's answer actually makes sense for this specific customer, not just customers in general. If you're using Luna to draft replies at volume, review a random sample each day — ten out of a hundred — to catch when it starts drifting off your tone or making stuff up. Keep a human in the loop for any message that could damage a relationship if it's wrong.