# OpenAI Decisions API preview: GPT-6 Luna's accuracy and confidence

> When GPT-6 Luna, the model behind OpenAI's Decisions API, says it is 99% or more sure, it is right 68.7% of the time on open-world problems and 68.1% on closed-world ones; Jev is right 99.4% and 98.5%. A preview built on Luna, not a test of the API itself.
