OpenAI Reduces Prices for GPT-5.6 Luna and Terra Models

OpenAI announced price reductions for its low and mid tier AI models on Thursday. The cost of the lightweight GPT-5.6 Luna model was cut by 80 percent, reducing input costs to 0.20 dollars per million tokens and output costs to 1.20 dollars per million tokens. The mid tier GPT-5.6 Terra model saw a 20 percent price reduction, bringing input costs to 2 dollars and output costs to 12 dollars per million tokens. The company did not change the price of its flagship Sol model, though it introduced a fast mode for Sol that can increase performance speed by up to 2.5 times.

OpenAI attributed the price cuts to efficiency gains within the GPT-5.6 system. These changes apply to API usage for developers and businesses, while subscriptions for ChatGPT and Codex remain unchanged. The move comes as corporate users express concern over ballooning AI costs and the unpredictability of usage based pricing.

Industry analysts suggest the pricing shift is a response to intensifying competition. OpenAI now undercuts the pricing of Anthropic's mid tier Claude Sonnet 4.6 model. Additionally, the company faces pressure from Chinese rivals such as Zhipu AI, MiniMax, and Moonshot AI, which offer low cost or open weight alternatives. According to Artificial Analysis, the price drop makes GPT-5.6 Luna one of the most attractive models in terms of intelligence per dollar.

Outlets with a center lean frame the event as a strategic response to business cost scrutiny and general industry competition. In contrast, a center right outlet frames the move as OpenAI blinking in a face off with Chinese rivals.