OpenAI Embeddings API Guide: Models, Dimensions & Pricing

Choose an embedding model

Model	Default dimensions	Custom dimensions	Verified price
`text-embedding-3-large`	3,072	Yes	$0.13 / 1M input tokens
`text-embedding-3-small`	1,536	Yes	$0.02 / 1M input tokens
`text-embedding-ada-002`	1,536	No	$0.10 / 1M input tokens

Use text-embedding-3-small for the lowest cost, or text-embedding-3-large when retrieval quality matters more. Keep one model and one vector size within an existing index.

Create embeddings

input accepts a string or an array. The array order is preserved in data[].index.

cURL · batch with reduced dimensions

curl https://api-models.com/v1/embeddings \
  -H "Authorization: Bearer $API_MODELS_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "text-embedding-3-small",
    "input": ["First document", "Second document"],
    "encoding_format": "float",
    "dimensions": 256
  }'

Python · requests

import os
import requests

response = requests.post(
    "https://api-models.com/v1/embeddings",
    headers={
        "Authorization": f"Bearer {os.environ['API_MODELS_KEY']}",
        "Content-Type": "application/json",
    },
    json={
        "model": "text-embedding-3-large",
        "input": ["First document", "Second document"],
        "encoding_format": "float",
        "dimensions": 1024,
    },
    timeout=120,
)
response.raise_for_status()
vectors = [item["embedding"] for item in response.json()["data"]]
print(len(vectors), len(vectors[0]))

Node.js · fetch

const response = await fetch("https://api-models.com/v1/embeddings", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.API_MODELS_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "text-embedding-3-small",
    input: ["First document", "Second document"],
    encoding_format: "float",
    dimensions: 256,
  }),
});
if (!response.ok) throw new Error(await response.text());
const result = await response.json();
console.log(result.data.map((item) => item.embedding.length));

Parameters and limits

Field	Requirement	Notes
`model`	Required	Use an exact model ID from the table
`input`	Required	String, token array, or batch; no empty strings
`encoding_format`	Optional	`float` or `base64`; default `float`
`dimensions`	v3 only	Shortens vectors from `text-embedding-3-large` or `text-embedding-3-small`
`user`	Optional	Stable end-user identifier for abuse monitoring

Each input is limited to 8,192 tokens. A request can contain up to 2,048 inputs, with no more than 300,000 total input tokens across the request.

Store and compare vectors correctly

Persist the model ID and dimensions with every index version.
Use the same embedding model for stored documents and search queries.
Normalize or choose a distance metric supported by your vector database.
Rebuild the index when you change model or dimensions; vectors from different spaces are not interchangeable.

Common errors

Symptom	Fix
dimensions rejected	Use a `text-embedding-3` model; ada-002 does not support it
Vector size mismatch	Set the database column/index to the returned dimension
Context length exceeded	Chunk the text below 8,192 tokens per input
Chat-style error	Use `/v1/embeddings`, not `/v1/chat/completions`