OpenAI announced significant price reductions and performance improvements for its GPT‑5.6 model family. GPT‑5.6 Luna now costs 80% less, while GPT‑5.6 Terra is 20% cheaper, making high‑volume AI work far more economical. These lower prices apply across Chat GPT Work, Codex, and the API, with Luna priced at $0.20 per million input tokens and Terra at $2 per million input tokens.

The company also introduced Fast mode for GPT‑5.6 Sol, delivering up to 2.5× faster speeds than Standard processing at twice the price, replacing the previous Priority Processing tier. Fast mode is backward compatible with existing priority‑tagged requests.

OpenAI emphasized matching intelligence to outcomes: Luna handles high‑volume, well‑specified tasks at roughly 6 cents on the dollar compared to year‑old frontier models, while Sol tackles complex, high‑stakes work. A coding workflow might use Sol for planning and Luna for implementation and testing.

Efficiency gains come from model improvements, optimized inference systems, and agentic harnesses. GPT‑5.6 Sol itself helped rewrite production kernels, cutting serving costs by 20% and boosting token‑generation efficiency over 15%. This feedback loop accelerates future improvements.