Overview
- OpenAI announced on Thursday, July 30, 2026, that GPT‑5.6 Luna will cost $0.20 per million input tokens and $1.20 per million output tokens and that GPT‑5.6 Terra will cost $2 per million input tokens and $12 per million output tokens while Sol’s standard price remains unchanged.
- The company attributed the reductions to model and infrastructure improvements such as rewritten GPU kernels, more efficient token generation, and greater prompt caching that cut serving costs and raised tokens‑per‑dollar.
- OpenAI adjusted how Luna and Terra usage counts against paid ChatGPT Work and Codex quotas so subscribers will consume fewer credits and can complete more tasks before hitting limits.
- Executives and analysts framed the move as a response to enterprise resistance to soaring usage bills and growing pressure from lower‑cost Chinese open‑weight models, and they warned the cuts could spur wider price responses across competitors.
- For businesses, the lower prices make agentic and high‑volume background work cheaper to run, but the change also tightens vendor margins and increases the need for firms to track token‑based billing and route work to the right model.