OpenAI has officially moved its Decisions API into general availability, offering a specialized service designed to bypass long text generation in favor of fast, structured choices. The API—powered by the GPT-6 Luna reasoning engine—processes up to 400 predefined questions per request while cutting typical response latencies down to roughly 100–180 milliseconds. The updated release adds native GPT-Live voice integration and image input modalities, targeting high-frequency agentic workflows, classification tasks, and real-time routing. Pricing is structured around input tokens (at $0.10 per million tokens, or $0.01 when cached) with zero charges applied to output tokens, directly challenging rivals in the low-cost decision market. Early adopters including Box, Speak, and Ramp report significant speed and efficiency gains, marking a sharp industry shift toward targeted, modular AI decision layers over general conversation models.

Full report : OpenAI’s Decisions API general release introduces native voice integration and image inputs to fight models like TypeSafe’s Jev.