The API is intended for near-real-time classification and evaluation of text and images without generating a full model response.
The announcement lists native GPT-Live integration for voice applications, prompt caching, native JSON inputs, and image instructions for individual questions. A request can now contain up to 400 questions, which OpenAI describes as twice the prior limit.
For gpt-6-luna on the Decisions API, OpenAI lists input pricing of $0.10 per million non-cached tokens and $0.01 per million cached tokens. It says there are no output-token charges; cached-write pricing for 30-minute and 12-hour periods is referenced separately in its pricing guide.
The source does not provide independent latency or quality measurements, comparative benchmarks, regional availability details, or evidence of production results from customers.