# Decisions models Decisions runs on `POST /v1/decisions`. Pass the model id in the `model` field. ## SPEV SPEV answers predicate, choice, and score questions about text, audio, or an image, with a probability for every candidate answer. | Field | Value | | --- | --- | | Model ID | `spev` | | Evidence | Text, audio, or one image per request | | Endpoint | `POST /v1/decisions` | | Audio | One clip as `input_audio` | | Image | One `input_image` part across all messages | One request can ask many questions about the same evidence. Send them together rather than one request per question. ## How it compares Several models read the same evidence. Pick by what you want back. | Model | Returns | Use it when | | --- | --- | --- | | SPEV | Typed answers with probabilities | Your code branches on the answer | | [Symphony Notes](/docs/captioning/models) | Prose describing a recording | You want a description to read or index | | [Chat models](/models) | Free-form text | You want an open-ended reply | See the [model catalog](/models) for the live list and [pricing](/pricing) for current rates.