OpenAI Slashes GPT-5.6 Model Prices to Fight Chinese Rivals

OpenAI reduced API pricing for two models in its GPT-5.6 family on August 1, 2026, to combat rising cost scrutiny from enterprises and competition from Chinese rivals. The company cut the cost of the fastest model, Luna, by 80% to 20 cents per million input tokens and $1.20 per million output tokens. The mid-tier Terra model saw a 20% reduction, bringing costs to $2 per million input and $12 per million output tokens. Pricing for the high-end Sol model remains unchanged, though a new Fast mode was introduced that offers 2.5 times faster speeds at twice the standard price.

These reductions stem from technical efficiency gains, including recursive improvements where the Sol model was used to optimize its own software kernels, resulting in 15% higher efficiency. The shift targets a transition from maximizing usage to prioritizing clear returns on investment, as businesses move toward a "valuemaxxing" approach. The move also responds to pressure from open-weight models like Moonshot AI's Kimi K3, DeepSeek V4, and Z.ai's GLM-5.2.

Simultaneously, the company launched ChatGPT for Academic Researchers, providing free access to the Sol Pro model for 10,000 researchers this summer, supported by a $250 million commitment. This strategy coincides with OpenAI reaching one billion monthly active users and two million business customers, while facing aggressive pricing from U.S. competitors including Google, Microsoft, and Anthropic.