community.openai.com via Reddit

OpenAI launches Decisions API public beta on GPT-6 Luna

By Published Updated
6 sources tracking this story
OpenAI Agents Inference ai-business

TL;DR

  • Billing output at zero cuts total Decisions API cost to roughly one-fifth of what standard Luna pricing implies for classification workloads where answers are short.
  • gpt-6-luna's 18-row endpoint support table omits /v1/decisions; only Chat Completions, Responses, and Batch are marked Supported.
  • The 10x speed claim over the Responses API ships without a benchmark or per-call compute disclosure in the public beta documentation.

OpenAI opened a dedicated Decisions API in public beta on October 6, running only on gpt-6-luna. The announcement on the OpenAI developer forum prices it at $0.10 per 1M input tokens and charges nothing for output tokens or for cache reads and writes.

The endpoint hits POST /v1/decisions and returns typed answers in three shapes: a predicate as a 0-to-1 probability, a choice from fixed options with confidence scoring, and a score rated against ordered severity levels. Text and images are both accepted, but as Unite.ai notes, image inputs must be inline base64 data URLs — hosted image links and uploaded file IDs are refused.

The pitch is speed. OpenAI says the endpoint "produces those answers about 10x faster than the Responses API." General availability is expected "in the coming weeks."

Alongside the price-per-token line, the beta ships with Zero Data Retention support, HIPAA eligibility, and data residency in the US and Europe (EEA/Switzerland). Those compliance checkboxes make it a plausible drop-in for moderation and gating flows that previously stitched together JSON-schema prompts on Chat Completions.

The 10x figure is measured against OpenAI's own Responses API, not an external vendor. Neither the forum post nor the follow-on coverage publishes a head-to-head accuracy number comparing /v1/decisions to the same gpt-6-luna model called through the Responses API.

What others are reporting

Coverage cluster as of 9h after publish

  1. OpenAI Docs Read →

    First-party reference confirming the three typed answer shapes (predicate, choice, score), gpt-6-luna exclusivity, and minimum SDK versions; positions GA as weeks away.

    The Decisions API evaluates text, images, or both and returns typed answers about 10x faster than the Responses API.
  2. MIXED Read →

    Flags the documentation inconsistency: /v1/decisions is absent from Luna's 18-row endpoint support table, where Chat Completions, Responses, and Batch are the only three marked Supported.

    v1/decisions is not among them. Chat Completions, Responses and Batch are the three marked Supported.
  3. OrcaRouter Read →

    Corrects early coverage that assumed standard Luna output rates apply; identifies missing beta items including answer-set limits, fine-tuning, and unconfirmed GA pricing.

    You pay for the evidence you send, not for the answer that comes back.
  4. Vercel Read →

    Covers production deployment via Vercel AI Gateway using the openai/gpt-6-luna-decisions model ID and experimental_decide in the AI SDK across routing, classification, and image-branching patterns.

    Decisions API gives that kind of choice its own interface.
  5. Unite.AI Read →

    Adds compliance detail (HIPAA, Zero Data Retention, US/EEA residency) and the base64-only image input constraint, plus Live API voice integration as a downstream use case.

    The new endpoint produces those answers about 10x faster than the Responses API.

Originally reported by community.openai.com

Read the original article →

Original headline: OpenAI Opens Decisions API Public Beta on GPT-6 Luna, Prices Input at $0.10 per Million Tokens