OpenAI has officially launched its Decisions API in public beta, introducing a dedicated endpoint designed for classification, choice selection, and calibrated probability scoring. Powered exclusively by the gpt-6-luna model, the API processes text and image inputs to return typed array responses without generating unstructured conversational prose.
According to OpenAI, the Decisions API operates roughly 10 times faster than the standard Responses API, reducing latency from 1.6 seconds down to approximately 150 milliseconds. The pricing model eliminates output token fees entirely, charging a flat $0.10 per million input tokens, though standard long-context and regional premiums still apply.
The release directly competes with existing decision models like TypeSafe’s Jev. While OpenAI’s API carries higher base input costs than competitors, it offers immediate integration advantages, including native multimodal image input, enterprise compliance guarantees, and open beta accessibility via existing developer tools.
Why it matters
Sub-200ms response times make LLM-backed decision logic feasible for real-time application routing and fast UI branching.
Zero output token pricing simplifies API cost forecasting for high-volume automated workflows that do not require text generation.
Multimodal support allows developers to run fast visual classification and decision routing within the same API endpoint.
Source: marktechpost.com



