GPT-6 Sol and Luna shipped 22 September. If you already run OpenAI agents, content pipelines, or Codex, unit cost just dropped. Luna is cheap enough to sit on inbound email and CRM notes without a finance meeting.
OpenAI priced GPT-6 Sol at $2 per million input tokens and $10 per million output. That is half of GPT-5.6 Sol’s promotional rate of $4 / $20. Luna is $0.10 / $0.50, down from $0.20 / $1.20. Cached input is 10% of the input rate. Both sit under Astra at $10 / $50, which stays the expensive flagship. Availability is Plus, Pro, Business, Enterprise, and Edu in ChatGPT Work and Codex. Free and Go get Luna on desktop only.
This is not a new vendor. It is a cheaper engine in a stack you already pay for. Sol is the mid-tier for coding and longer jobs. Luna is the high-volume workhorse: classify, summarise, tag, draft the first reply. For an agency or a five-person product shop, the win is moving volume off Sol and Astra onto Luna so the expensive model only touches work a human will actually review.
OpenAI will talk quality. Third-party writeups say the price move is larger than the quality move versus 5.6. Treat “half the factual mistakes” as marketing until you run your own tickets. The number you can bank is the invoice. Astra is still the wrong default for inbox triage.