An 80% price cut changes which AI tasks are worth automating

Subject: An 80% price cut changes which AI tasks pay

Preview: Token prices fall. Retries, controls and exceptions still determine the real cost.

OpenAI announced by email on 31 July that it was cutting GPT-5.6 Luna pricing by 80% and Terra pricing by 20%. If those rates are active for your account, agents that classify, monitor or enrich thousands of records may become far cheaper to run.

The official email lists Luna at $0.20 per million input tokens and $1.20 per million output tokens, down from $1 and $6. Terra would move to $2 input and $12 output. A fast option for Sol promises up to 2.5 times the speed at twice the standard price.

The larger opportunity is background work: continuous monitoring, extraction, quality checks, case preparation and classification. A small saving per call becomes material when a workflow performs millions of actions.

Token price is not cost per accepted outcome. Retries, errors, tool calls, human review and exceptions can erase the advantage. The public rate pages we checked still displayed the earlier prices, so confirm the live rate card before building a budget.

Measure cost per correctly completed case. Route difficult decisions to Sol, routine production to Terra and repetitive volume to Luna. Test on real cases with an explicit quality threshold.

Cheaper AI creates value only when the complete outcome becomes cheaper.