The OpenAI Decisions API is GA with documented batching, pricing and data controls.
OpenAI makes the Decisions API generally available
2026-10-22Verified announcementTRIAL
OpenAI moved `/v1/decisions` from public beta to GA. Using GPT-6 Luna, the API returns probabilistic predicate, choice or score answers over text, images or JSON. The per-request limit rises from 200 to 400 questions, with prompt caching, usage reporting, Private Link and regional processing. Uncached input costs $0.10 per million tokens and output tokens are free.
What this means for you
For routing, review and classification teams, GA status, batched questions and caching make the specialized endpoint a production-comparison candidate. OpenAI claims roughly 10x the speed of the Responses API, but thresholds, calibration error and mistake cost require validation on local labels.
Evidence & related material
Technical details & scores
| Stage | GA |
|---|---|
| Endpoint | POST /v1/decisions |
| Model | gpt-6-luna only |
| Inputs | text, images, JSON, conversation and tool history |
| Outputs | predicate probability, discrete choice distribution or ordered score |
| Request Limit | 1–400 questions |
| Price | $0.10 per 1M uncached input tokens; output tokens free; cache, region and long-context rates apply |
| Data Controls | ZDR and HIPAA for eligible customers; residency in supported regions; US/EU regional processing |
| Vendor Claim | about 10x faster than Responses API; not independently verified |
| Time Basis | official changelog gives date only; strict hourly placement is unavailable |
Technical impact 3
Engineering value 5
Adoption signal 3
Maturity 4
Verification cost 2
Risk 3