Overview
- OpenAI moved the Decisions API to general availability on October 22, 2026, making the endpoint accessible to eligible developers in select global markets.
- The service runs on GPT-6 Luna and uses an input-only pricing model that charges $0.10 per million uncached input tokens and $0.01 per million cached input tokens while not charging for output.
- Between the public beta and GA OpenAI raised the per-request limit to 400 questions and added native GPT-Live integration, prompt caching, and support for images and JSON inputs.
- OpenAI showcased speed gains for bounded decision tasks, saying the Decisions API can be up to 10 times faster than calling Luna through the Responses API with demonstration latencies around 150 milliseconds, though those figures come from vendor demos and may vary in practice.
- The API includes enterprise options such as Zero Data Retention, HIPAA eligibility, and US/EU inference residency, and it is aimed at routing, classification, and other micro-decision use cases where compact structured outputs cut compute and cost needs.