OpenAI expanded GPT-6 with Sol and Luna on September 22, 2026, about 90 minutes after Anthropic's Opus 5.5 drop. TechCrunch reports Sol targets complex work like coding, while Luna targets high-volume clerical tasks. The story is less about a new peak model and more about cutting the bill for work you already do.
API prices land at about half of GPT-5.6 Sol and Luna. GPT-6 Sol is $2 input and $10 output per million tokens. Luna is $0.10 input and $0.50 output. OpenAI attributes the drop to caching and inference improvements. On internal factuality evals, the company claims GPT-6 Sol makes about half as many mistakes as its predecessor and reaches Astra-level reliability at lower cost, with a lower coding error rate as well. Ars Technica called it an iterative, cost-focused upgrade rather than a new frontier peak, which matches how most production teams will use it.
Availability is broad for paid users. Sol shows up in ChatGPT Work and Codex for most paid accounts and via the API. Luna is in the desktop app for Free and Go, with a gradual ChatGPT app and web rollout. That split makes sense if Luna is the high-volume clerk and Sol is the heavier coding partner.
If you run agents or CI helpers, Sol's half-price API and fewer mistakes are the numbers to model first. If you push a lot of form-filling, summarization, or other clerical traffic, Luna's $0.10 / $0.50 pricing is designed for that volume. Treat company factuality claims as internal until you rerun your own eval suite. The public story from TechCrunch and Ars is clear on price and product split, not on independent leaderboard crowns.
If you already budgeted for GPT-5.6 Sol or Luna, cut those line items roughly in half and re-run your latency and error dashboards after you switch. A cost cut this large only helps if quality holds on your prompts, not just on OpenAI's internal factuality set. Pair Sol with the coding paths that used to burn tokens, and park Luna on the repetitive clerical queues where cheap volume wins.
I want models that are cheap enough to leave on in the background without a finance review every Friday. Sol and Luna lean that way: same GPT-6 family, sharper pricing, and a role for both deep coding and mass clerical work. That is the kind of release that shows up in next month's invoice.
