Cheaper AI Does Not Mean Cheaper Work
Luna's API price fell 80%, but list price is only one part of operating cost. Measure accepted outputs, review time, retries, integration, and operating complexity before you move a workflow.
In this edition
The decision in 30 seconds
Read this if you read nothing elseReprice one repeatable workflow on ten representative jobs and count only outputs that meet the quality bar.
Luna fell to $0.20 per million input tokens and $1.20 per million output tokens, one-fifth of its prior list price.
Routine, high-volume work became cheaper at list price, but the operating question is still cost per accepted result after review and retries.
Do not migrate from list price or provider benchmarks alone. A cheaper run can still cost more when quality, review, retries, integration, or operating complexity gets worse.
Do this week
Only where it appliesReprice one real task
Run the same ten representative jobs on your current model and Luna. Count only outputs that meet the quality bar, then include review and retry time.
Limit: Do not move customer-facing, regulated, or high-cost-of-error work on list price alone.
The cost curve moved again
Relevant if API usage or agent credits are a material operating cost.
PricingCompany pricing; current API documentationOpenAI cut Luna by 80% and Terra by 20%Luna moved from $1 input and $6 output per million tokens to $0.20 and $1.20. Terra moved from $2.50 and $15 to $2 and $12.Applies to: Teams running high-volume, repeatable API or Codex workflows
For a workflow using 10 million input tokens and 2 million output tokens, Luna's list-price cost falls from $22 to $4.40, a saving of $17.60. OpenAI kept ChatGPT and Codex subscription prices and quota budgets unchanged, but says Luna and Terra now consume fewer credits.
Reprice one high-volume workflow, then divide the full cost by outputs that met your quality bar. A price falling to one-fifth is worth testing, not an automatic migration. Long prompts above 272,000 input tokens carry higher rates.
Sources: OpenAI, OpenAI API docs
Sources and further reading
Evidence note: Sources checked August 14, 2026. OpenAI's performance and cost comparisons are company claims. Published API prices are list prices and do not include review time, retries, integration, or operating complexity.
Get the next Skim.
New editions by email when they clear the evidence and usefulness checks.
Bring one decision. Leave with a next step.
Use the 30-minute call to work through that specific model, workflow, or risk choice.