Decisions API
A decision-oriented API interface introduced by OpenAI, built on the Luna model and designed for developer scenarios that require AI-assisted decision-making.
Core Facts
The Decisions API runs on a model called GPT-6 Luna
OpenAI announced that the Decisions API is open for public testing to all developers, powered by GPT-6 Luna and supporting text and image input.
OpenAI launched a product called the Decisions API.
Timeline (last 90 days)
Starting from TypeSafe's release of Jev in mid-September 2026, at least eight companies launched AI decision models within a little over a month, including Cloudflare's Clef, the AWS Strands team's Decider, Perplexity's pplx-decider, Liquid AI's d1 series, OpenAI's Decisions API and Microsoft-Decision-1.
OpenAI Decisions API images must be passed in as inline base64, and hosted URLs and file IDs are both rejected.
OpenAI officially claims the Decisions API is about 10 times faster than the Responses API, but it did not disclose benchmarks and did not publish head-to-head accuracy comparisons with competitors.
OpenAI released the public beta of the Decisions API, with the endpoint POST /v1/decisions, which returns only typed answers (probability, choice, or score).
The OpenAI Decisions API offers zero data retention support and HIPAA compliance eligibility in the beta.
On October 6, 2026, OpenAI posted an announcement on its developer forum announcing that the Decisions API had entered public beta.
One analysis concluded that in short-answer classification scenarios, the cost of the Decisions API is about one-fifth of standard gpt-6-luna pricing.
OpenAI's documentation mentions two variables for the Decisions API—regional processing surcharges and a long-context input multiplier—but neither has published specific numbers.
The Decisions API is still in public beta, and general availability (GA) will launch within the coming weeks, with no fixed date.
The Decisions API supports Zero Data Retention, is HIPAA eligible, and offers US/EU data residency options.
40 more timeline events
All Facts (20)
The Decisions API runs on a model called GPT-6 Luna
70%VerifiedOpenAI announced that the Decisions API is open for public testing to all developers, powered by GPT-6 Luna and supporting text and image input.
70%VerifiedOpenAI launched a product called the Decisions API.
75%UnverifiedOpenAI officially claims the Decisions API is about 10 times faster than the Responses API, but it did not disclose benchmarks and did not publish head-to-head accuracy comparisons with competitors.
50%UnverifiedThe OpenAI Decisions API offers zero data retention support and HIPAA compliance eligibility in the beta.
50%UnverifiedOpenAI Decisions API images must be passed in as inline base64, and hosted URLs and file IDs are both rejected.
50%UnverifiedOne analysis concluded that in short-answer classification scenarios, the cost of the Decisions API is about one-fifth of standard gpt-6-luna pricing.
50%UnverifiedThe Decisions API currently supports only one model, gpt-6-luna, with no fallback model option.
50%UnverifiedOpenAI's documentation mentions two variables for the Decisions API—regional processing surcharges and a long-context input multiplier—but neither has published specific numbers.
50%UnverifiedThe Decisions API supports Zero Data Retention, is HIPAA eligible, and offers US/EU data residency options.
50%UnverifiedDecisions API input can be text, images, or a mix of both, but images must be provided as base64 inline data URLs (up to 128 per request), and hosted image links and uploaded file IDs are rejected.
50%UnverifiedDecisions API output tokens are billed at $0, and billing is calculated on input only.
50%UnverifiedOpenAI's official guidance draws three lines: the Decisions API for classification, routing, scoring, and gating that require a determination; Structured Outputs (Responses API) for scenarios that require extracting fields or written explanations in a structured form; and function calling for scenarios where the model triggers tools with parameters.
50%UnverifiedThe OpenAI Decisions API (gpt-6-luna) was released on October 6, with an input price of $0.10 / 1M tokens, free output, closed source, and an undisclosed context length.
50%UnverifiedThe Decisions API returns three typed answer formats: predicates (yes/no with probability), choices (select one from a fixed list), and scores (rated against a rubric), without generating long-form text.
50%UnverifiedIn a marble-jar experiment where the true probability of drawing red was 50%, the Decisions API chose red 86% of the time; moving the red option from first position to last reduced that probability to 73%.
50%UnverifiedThe decision-API category only makes sense at scale (millions of content-moderation calls, enterprise-level ticket routing, high-consumption agent loops); at small scale, an ordinary model with a JSON schema can accomplish the same task.
50%UnverifiedWhen using decision APIs, scores should be treated as hints rather than measurements; build your own thresholds, test answer-option ordering before going live, and note that the 10× speed claim only holds when the model was previously generating text that wasn't actually needed.
50%UnverifiedThe order of answer options causes the Decisions API's confidence to drift by approximately 13 percentage points.
50%UnverifiedLangWatch's LLM-as-judge benchmark results show: OpenAI Decisions at 75.2% accuracy, 243 ms median latency, and $1.09 per 1,000 calls; TypeSafe Jev at 63.7% accuracy, 429 ms median latency, and $0.42 per 1,000 calls.
50%