#openai#decisions-api#gpt-6-luna#automatizacion

    OpenAI Decisions API: classify tickets without writing answers, pricing and limits

    OpenAI Decisions API: classifies text and images with GPT-6 Luna at $0.10 per million input tokens. What it returns, limits and beta.

    OpenAI Decisions API is OpenAI's new API that does not write answers: it only decides. On October 6, 2026 it launched in public beta (POST /v1/decisions) and, instead of text, returns typed answers: a probability, a category with its confidence, or a score against levels you define. It accepts text and images and, according to OpenAI, is about ten times faster than the Responses API. With GPT-6 Luna, input costs $0.10 per million tokens.

    What the Decisions API is

    To classify a customer message, the AI does not need to write anything. Picking a label is not the same as writing an answer, and this API is built for the former. Instead of a paragraph you then have to interpret, you get a value your system can use directly.

    The three things it decides

    1. Probability. It calculates how likely it is that something is true.
    2. Category with confidence. For example, it picks a department for a message and says how sure it is.
    3. Score against levels. It scores a text against levels you define.

    Customer support agents with headsets and laptops, the kind of classification work the Decisions API can automate

    What it costs

    What almost nobody mentions: with GPT-6 Luna, input costs $0.10 per million tokens, and output, cache reads and cache writes are not charged. You only pay for what you send. It also understands images. For GPT-6 context in ChatGPT, read ChatGPT Intelligent UI with GPT-6.

    The fine print

    • It only works with gpt-6-luna for now.
    • It is still in beta: OpenAI expects general availability in the coming weeks.
    • It does not accept URLs or file_id: images go in as base64.

    How to apply it in your business?

    Instead of reading a thousand tickets by hand, you can review only the low-confidence ones. That use is a WAiBOT proposal, not an OpenAI feature. To start:

    1. Pick a message you classify by hand: tickets, emails or forms.
    2. Decide which question you ask it: probability, category or score.
    3. Send the text to the API with gpt-6-luna.
    4. Manually review only the cases where confidence is low.
    5. Measure how much time you save before scaling up.

    Frequently asked questions

    What does the Decisions API return?

    Typed answers: a probability, a category with its confidence or a score against levels you define.

    How much does it cost?

    With GPT-6 Luna, $0.10 per million input tokens. Output, cache reads and cache writes are not charged.

    Which models does it support?

    For now only gpt-6-luna.

    Is it available to everyone?

    It has been in public beta since October 6, 2026. OpenAI expects general availability in the coming weeks.

    Conclusion

    The OpenAI Decisions API changes the question: instead of asking the AI to write, you ask it to decide and you pay only for what you send. Confirm prices and limits in OpenAI's guide and pricing page before integrating it.