OpenAI just made its most aggressive pricing move yet. The company’s new GPT-5.6 model family, launched on July 9, introduces a three-tier system where the cheapest option, Luna, costs $1 per million input tokens and $6 per million output tokens. The top-tier Sol model in the same family runs $5 input and $30 output per million tokens. The middle tier, Terra, sits at $2.50 input and $15 output, reportedly matching GPT-5.5 performance at approximately half the price.
The three-tier play and what it actually costs
At $1 per million input tokens, OpenAI claims Luna achieves roughly 1/16th the cost of competitors like Claude Fable 5. The claimed 25x cost reduction appears to reference comparisons against older flagship models or competitor pricing rather than a strict apples-to-apples comparison within OpenAI’s own latest lineup. The practical reduction within the GPT-5.6 family itself is closer to 5x between Luna and Sol.
Health intelligence is the quiet headline
OpenAI has been steadily building out health and wellness capabilities, a focus that was reinforced during the GPT-5.5 enhancement cycle in June 2026. By that point, GPT-5.5 was already handling over 230 million weekly health-related queries. At Luna’s price point, continuous monitoring applications become economically rational. At $5 per million input tokens, the math was uncomfortable. At $1, it starts to pencil out.
