OpenAI launches beta Decisions API with GPT-6 Luna for typed classifications
The new endpoint returns predicates, choices and rubric scores from text or images. OpenAI claims roughly 10× lower latency than the Responses API, but has not published the benchmark methodology.
2 min read
OpenAI released the Decisions API in public beta on October 6, giving developers a dedicated endpoint for classification and scoring work. The new POST /v1/decisions route currently supports only GPT-6 Luna and accepts text, images or both.
Instead of generating free-form prose, the API returns one of three typed answer formats: a predicate with a probability from 0 to 1, a choice from developer-supplied options, or a score against ordered levels. OpenAI positions it for routing requests, classifying content and prioritizing work where parsing a conventional model response would add latency and failure modes.
The company says Decisions is about 10 times faster than using the Responses API for comparable tasks. That is a vendor-reported performance claim, not a published benchmark: the documentation does not disclose a test set, request sizes, regions, concurrency, latency percentiles or exact Responses API settings. Teams should therefore measure speed, calibration and decision quality on their own production distributions before replacing deterministic rules or specialized classifiers.
GPT-6 Luna remains a hosted, proprietary API model rather than a downloadable weight release. Its model page lists text input and output, image input, a 1.05-million-token context window and a 128,000-token maximum output. Standard pricing is listed at $0.10 per million input tokens, $0.01 per million cached input tokens and $0.50 per million output tokens, with different rates for long prompts and processing tiers. The Decisions guide’s example reports zero output tokens, but OpenAI’s public documentation does not yet provide a separate endpoint-specific billing explanation.
The beta is available in supported API regions, with regional processing offered in the United States and Europe. OpenAI says general availability is expected in the coming weeks, so endpoint behavior and limits may still change.
Sources: OpenAI Decisions guide · API changelog · GPT-6 Luna model page · OpenAI data controls · Unsplash image license
Conversation
0 approved commentsNo comments yet. Start a thoughtful conversation.
Join the conversation.
Sign in with ChatGPT to leave a comment or rate this article.