OpenAI has rolled out its Decisions API in public beta, pricing it aggressively at ten cents per million input tokens with output tokens thrown in for free. It reads as a direct maneuver to capture the unglamorous corporate plumbing market.
According to OpenAI, the infrastructure runs roughly ten times faster than the standard Responses API. It handles three distinct response formats: binary probabilities, predefined categories, and scale-based ratings for both text and images.
As OpenAI stated, practical applications include routing customer inquiries, classifying documents, and parsing damage in photos without the usual overhead. For corporate buyers, the setup supports zero-data retention and HIPAA compliance across the US and Europe.
It looks like a defensive play against the mini-trend of specialized decision models that emerged in mid-September. Instead of letting smaller players siphon off parsing budgets, OpenAI is undercutting the cost of routine classification.
Alongside the API drop, OpenAI pruned its paid tiers from five down to three: Build, Launch, and Grow. Organizations jump tiers automatically once cumulative credit purchases hit the designated threshold. Monthly usage caps now sit at $500 for Build, $5,000 for Launch, and $200,000 for Grow, forcing teams to plan their compute spend with rather less ambiguity than before.