OpenAI cuts GPT-5.6 Luna and Terra prices and accelerates Sol
OpenAI reduced GPT‑5.6 API prices on July 30, 2026: Luna became 80% cheaper and Terra 20% cheaper. Luna now costs $0.20 per million input tokens and $1.20 per million output tokens; Terra costs $2 and $12 respectively.
Why lower inference costs matter
The cuts make high-volume agent workflows, document analysis, customer-interaction classification and routine coding more economical. Luna supports tool use and multi-step execution, so the saving applies to completed processes rather than isolated text generation.
The new economics build on GPT-5.6 efficiency gains from lower token use, but the impact is operational: companies gain more freedom to match model capability to the value and risk of each task. OpenAI says Luna can deliver performance comparable to models that were frontier-class a year ago at roughly 6% of their cost per task and nearly nine times the speed.
Pricing, speed and infrastructure gains
OpenAI also replaced Priority Processing with Fast mode for GPT‑5.6 Sol. It provides up to 2.5 times the speed of Standard processing at twice the price, without changing model intelligence. Existing API requests tagged as priority automatically move to Fast mode, while standard Sol pricing remains unchanged.
The company attributes the gains to model routing, production software, context management and more efficient inference. Within a human-led process, Sol optimized production kernels and ran hundreds of experiments. The kernel work reduced end-to-end serving costs by 20%, while the experiments improved token-generation efficiency by more than 15%.
“GPT‑5.6 Terra is a strong fit for everyday work in Notion’s personal agent. In our evaluations, it delivered comparable quality to GPT‑5.5 at half the cost per task and in 60% less time.” — Hoda Noorian, Notion
Terra and Luna remain available through ChatGPT Work, Codex and the OpenAI API. Subscription prices and quota budgets have not changed, although use of the two models now consumes fewer credits.
For businesses, the practical move is to evaluate cost per completed task rather than token price alone. Sol can handle ambiguity and planning, while Luna executes defined steps, tests and routine automation; Terra fits everyday work where balanced quality and latency matter.