OpenAI launches Decisions API for bounded agent answers

OpenAI has made its Decisions API generally available, running on GPT-6 Luna to return bounded answers instead of generated text. Pricing is $0.10 per million input tokens, $0.01 cached, with no output token charges, and up to 400 questions per request.

• Three answer types: predicate probability, choice from options, and rubric score
• Up to 10x faster than asking GPT-6 Luna via the Responses API
• Up to 400 questions per request, with prompt caching and US/EU residency
• Plugs into GPT-Live as a trigger on transcript events in voice sessions