---
title: "Get the embedding vector for a paragraph"
method: GET
path: "/embeddings/{ref}"
tags: ["Embeddings"]
---

# Get the embedding vector for a paragraph

`GET /embeddings/{ref}`

Returns the embedding vector for a single paragraph.

Use `?model=large` (default) for the 3072-dimensional `text-embedding-3-large` vector — the canonical embedding used by the cross-references feature. Use `?model=small` for the 1536-dimensional `text-embedding-3-small` vector that powers `/search/semantic`.

The response includes `model` and `dimensions` fields so consumers can detect mismatches if they store vectors locally and compare against new responses. The `X-Embedding-Model` response header carries the same signal for byte-streaming clients.

## Path parameters

- `ref` string, required

## Query parameters

- `model` 'small' | 'large'

## Response `200`

Embedding vector

- object
  - `data` object, required
    - `standardReferenceId` string, required
    - `model` string, required
    - `dimensions` integer, required
    - `embedding` number[], required

## Other responses

- `404` — Paragraph or embedding not found

---

[API](https://skmtc.dev/urantia/apis/urantia-papers-api.md) · [All operations](https://skmtc.dev/urantia/apis/urantia-papers-api/llms.txt) · [OpenAPI document](https://skmtc-service-production.skmtc.workers.dev/v1/apis/urantia/urantia-papers-api/revisions/6a8c13929fda/schema)