OpenAI cuts GPT-5.6 prices and outlines a full-stack efficiency strategy

OpenAI said GPT-5.6 Luna pricing fell 80% to US$0.20 per million input tokens and US$1.20 per million output tokens, while Terra fell 20% to US$2 and US$12. It also described Fast mode for Sol as up to 2.5 times standard speed at twice the price, framing model, systems and infrastructure optimisation as one economic loop.

Impact: Lower inference prices change routing economics for high-volume agents and RAG. Architects should rerun quality-cost-latency benchmarks rather than assume last quarter’s model mix remains optimal.