OpenAI slashes GPT-5.6 Luna prices 80% in AI price war

OpenAI is escalating the AI price war. CEO Sam Altman announced major pricing cuts for the GPT-5.6 model family on July 30, 2026. GPT-5.6 Luna, the more affordable tier, now costs $0.20 per million input tokens and $1.20 per million output tokens — an 80% reduction. GPT-5.6 Terra was reduced by 20%, now at $2 per million input and $12 per million output. The top-tier GPT-5.6 Sol isn't getting a price cut; instead, OpenAI added a Fast mode in its API that delivers up to 2.5x speed at 2x the price, with no loss in intelligence. Altman framed the move as offering "the best combination of price and intelligence at every level." OpenAI attributes the efficiency gains to upgraded inference systems, smarter request routing, and optimized production software that reduces unnecessary compute iteration.

The pricing shift signals a broader industry transition. EMARKETER analyst Jacob Bourne called it the end of "tokenmaxxing," where businesses burn through tokens without equivalent value. Gartner's Arun Chandrasekaran said the cuts favor enterprise customers, as major providers move away from rigid pricing toward flexible commercial models. Competitive pressure is mounting from Moonshot AI's open-weight Kimi K3, while Google and Microsoft are also promoting cost efficiency — Microsoft's MAI-Thinking-1 model separates memory and tools from the core model to reduce inference costs. With OpenAI having reportedly filed confidentially for an IPO, the price war tests whether AI companies can sustain profitability while delivering demonstrable ROI to business clients.