OpenAI opens its Decisions API in public beta, built on GPT-6 Luna

OpenAI's Decisions API is a dedicated endpoint, POST /v1/decisions, that evaluates text, images or both and returns typed answers instead of prose. The docs say it is in public beta and expected to reach general availability in the coming weeks. The only supported model is gpt-6-luna.

There are three answer types. A predicate gives the probability that a condition is true. A choice picks one of your values with per-option probabilities and a confidence value. A score gives a probability-weighted average over ordered levels.

The docs price it at $0.10 per million input tokens. There are no output, cache-read or cache-write charges. Images go in as base64 data URLs or public URLs, not file IDs. Zero Data Retention and HIPAA use are supported for eligible customers, and a Playground is available. The docs say it is about 10x faster than the Responses API, without a workload definition.

Press coverage places an earlier limited preview at DevDay on 29 September and dates the public beta to 6 October. The date here rests on one secondary report, since the docs page carries no date.

The pitch is speed and cheap input. The proof is your own labelled examples, which is what OpenAI recommends for setting thresholds.

Sources

• https://developers.openai.com/api/docs/guides/decisions
• https://letsdatascience.com/news/openai-opens-decisions-api-public-beta-b69a1e40

Editor's first pass (AI-assisted)

• Real 4: Documented endpoint with public pricing and a Playground; labelled beta.
• Useful 4: Typed answers from text and images suit classification and routing without parsing prose.
• Notable 4: The largest lab now sells a decision endpoint, in a category others opened.
• Novel 3: It wraps an existing model behind a new endpoint; the interface idea predates it.