OpenAI cuts GPT-5.6 Luna price by 80%, Terra by 20%

OpenAI's price cuts come a day after the company described how GPT-5.6 helped make itself more efficient to run. The company credits improvements across models, inference systems, and the agentic harness for reducing the time, tokens, and cost required for each result. GPT-5.6 Sol played a key role in this process by autonomously rewriting and optimizing production kernels, which helped reduce the end-to-end cost of serving the model by 20%, and by running experiments that increased token-generation efficiency by more than 15%.

These efficiency gains are now being passed to customers, reflecting OpenAI's stated mission to make advanced intelligence more abundant and affordable. The strategy aims to serve both ends of the price-performance curve: lower-cost models like Luna make high-volume work economical at scale, while Fast mode gives API customers quicker access to Sol when response time is critical. Businesses can now define the outcome and quality standard they need, then choose between models like Sol for complex planning and Luna for well-specified implementation tasks.

The price cuts' real impact likely hinges on whether enterprises actually shift high-volume workloads to Luna and Terra at the new rates. If Luna performs on par with year-old frontier models at such a steep discount, more businesses may find it practical to run AI across broader operations like document analysis and customer-interaction classification. The feedback loop OpenAI describes—where more capable models help improve efficiency further—suggests that today's reductions may not be the last, should the company continue to find compounding gains.