AI Agent StoreOpenAI · Generally available Decisions API
GPT-6 Luna Decisions
Answer yes/no questions, choose named options or score records against your rules. Shared documents can be reused across requests, and each question can include its own images.
Released 2026-10-06 · First reviewed 2026-10-07 · Checked 2026-11-03.
We still need comparable independent tests. You can compare the confirmed features and limits below, but we cannot yet give this version an Agent Fit or Practical Value rating. Test it on records with known answers, including cases that should go to a person.
What you can use today
Token prices measure the small pieces of text the model reads or produces. Capacity is the amount the service accepts in a request; check the limits below for any fallback to another model. A dash means the rate is unverified.
Exact Decisions API model ID: gpt-6-luna. Generally available October 22. Send 1–400 questions with text, images or structured JSON. Public image URLs and base64 are supported; file_id is not. Question images share the request image limit. Context and effort limits are not established by this guide.
A generally available service using GPT-6 Luna, not a new base model. Chat benchmark results and vendor speed claims do not establish decision accuracy or measured latency here.
At least 1,024 identical shared-input tokens are needed for caching. Question text and question-specific images are outside the reusable prefix. Cache reads/writes are already included in reported total input tokens; do not add them again.
Output: Typed predicates, choices and rubric scores with probabilities and confidence fields. Compare decision models, limits and input costs ↗
Per million tokens: $0.10 uncached input, $0.01 cache reads, $0.125 standard cache writes or $0.25 for 12-hour writes. Answers are free. Region and long-context premiums still apply. Include retries when comparing the total cost.
Provider release and specifications ↗
Check availability and plans ↗
Test it before switching
Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. Check whether confidence scores match the answers you can verify.
Compare caching and long-context costs ↗