What happened
Pricing and supported-model details checked on October 28, 2026. This is an illustrative calculation, not a measured bill. SourceVane has not made live API calls or run a performance benchmark for this article.
Article
OpenAI made its Decisions API generally available on October 22, 2026. It is designed for bounded answers such as categories, probabilities and rubric scores, rather than open-ended writing. As checked on October 28, the developer guide lists gpt-6-luna as the supported model. [S1, S2]
For uncached input, the published base rate is US$0.10 per million tokens, with no output-token charge. Regional processing premiums and long-context multipliers can apply; cache writes also carry charges. These are Decisions endpoint terms, not a price for every request using the same model. [S2]
Illustrative calculation: 100,000 requests at 500 billed uncached input tokens each produce 50 million tokens. At the base rate, that is US$5. This assumes no cache writes, regional premium or long-context multiplier and excludes other services. It is arithmetic, not a measured bill or a savings claim.
Editorial recommendation: pilot one low-risk routing task with a labeled test set. Compare incorrect routes, review workload, latency and total billed cost against the current process. Keep human review for consequential or uncertain results. SourceVane has not run this benchmark.