OpenAI announced on Thursday significant price cuts for two of its newly released GPT-5.6 artificial intelligence models, Terra and Luna. The adjustment comes just three weeks after their public launch, signaling an aggressive push to retain enterprise customers who are increasingly scrutinizing AI expenditures.
The company is reducing the price of its mid-tier model, Terra, by 20%, bringing costs to $2 per million input tokens and $12 per million output tokens. The price reduction for Luna, the company’s fastest model, is more substantial at 80%, now priced at 20 cents per million input tokens and $1.20 per million output tokens. The flagship model, Sol, remains at its original price point.
This strategic shift occurs as businesses move away from the initial “tokenmaxxing” phase—where companies encouraged widespread, unrestricted AI use—toward a more cautious approach focused on return on investment (ROI). As AI operational costs for large organizations have reached billions of dollars, developers are facing intense pressure to provide more efficient, cost-effective solutions.
The competitive landscape has been further complicated by the emergence of high-performing open-weight models from Chinese startups. Notably, Moonshot AI recently released “Kimi K3,” which has demonstrated performance metrics that challenge leading American proprietary models. In response, rivals such as Anthropic have introduced the cost-efficient “Claude Opus 5,” and major players like Google and Microsoft have pivoted their recent product roadmaps to emphasize affordability and efficiency per task.
OpenAI stated that these price cuts were made possible by efficiency gains in the underlying infrastructure serving the models, framing the move as part of a long-term goal to make advanced intelligence more accessible and sustainable for enterprise deployment.