OpenAI just fired the loudest shot in August’s model price war. GPT-5.6 Luna — its fastest, cheapest model — dropped 80% to $0.20 per million input tokens, $1.20 output. Same model, one-fifth the price. At the same time, ChatGPT free users get Luna as their default model, with unlimited text chats and a new “Think” button for harder questions. ChatGPT now counts roughly 1 billion weekly users.
Where the 80% comes from
Not subsidy — engineering. Improved speculative decoding lifted token-generation efficiency by 15%+, and OpenAI says its own model rewrote production GPU kernels, cutting serving costs another 20%. Those savings pass straight through to the API, Codex, and ChatGPT.
The API angle
Luna at $0.20/1M input undercuts most hosted Chinese open-source models — the “open weights are cheaper” narrative just lost its headline number. For builders, Luna becomes the obvious default for high-volume work: classification, extraction, summarization, agent sub-tasks that don’t need a frontier brain. Anything you priced on Terra or Sol is worth re-benchmarking on Luna, because GPT-5.6 Terra also got 20% cheaper in the same move.
You Might Also Like
- Openai Oauth Turns Your Chatgpt Subscription Into a Free Openai api but Should you use it
- Openai Desktop Superapp Chatgpt Codex and Atlas are Becoming one app
- Openai gpt Realtime 2 Translate Whisper Three Voice Models one api Several Startups Erased
- Openai gpt 5 6 sol Terra Luna a Three Tier Lineup Only 20 Orgs can Touch
- Openai Sunsets Atlas Browser Agentic Browsing Folds Into Chatgpt Codex
Top AI Product 
Leave a comment