Embeddings

POST /v1/embeddings follows OpenAI’s shape. Embedding models are always explicit: there is no auto for embeddings. The key needs the embeddings scope (keys have it by default).

For the exact shape of every request and response, with curl, Python and Node examples, see the API reference.

Models

Model	Vendor	Max input tokens	Price (USD per million input tokens)
text-embedding-3-small	openai	8192	0.02
text-embedding-3-large	openai	8192	0.13
gemini-embedding-001	google	8192	0.20
mistral-embed-2312	mistral	8000	0.10

mistral-embed-2312 returns 1,024 dimensions and is processed by Mistral AI’s global API, which does not guarantee a location. Anthropic has no embedding models, and Together AI offers none on its serverless API. Prices are provider list prices; the plan fee applies on top (see Billing).

Request

curl https://api.zabio.ai/v1/embeddings \
  -H "Authorization: Bearer insert your zabio.ai key here" \
  -H "Content-Type: application/json" \
  -d '{"model": "text-embedding-3-small", "input": ["first text", "second text"]}'

Field	Notes
model	required; one of the ids above
input	a string or an array of 1 to 2048 strings
dimensions	optional; passed to the vendor
encoding_format	float (default) or base64 (little endian float32)
user	accepted

Response

{
  "object": "list",
  "data": [{"object": "embedding", "index": 0, "embedding": [0.0123, ...]}],
  "model": "text-embedding-3-small",
  "usage": {"prompt_tokens": 8, "total_tokens": 8},
  "zabio": {"vendor": "openai", "model_served": "text-embedding-3-small", "cost_micros": 0, "tokens_estimated": false}
}

zabio.tokens_estimated is true for Google models, which do not report token counts; Zabio estimates them from the input length and bills on the estimate.

Python

from openai import OpenAI

client = OpenAI(base_url="https://api.zabio.ai/v1", api_key="insert your zabio.ai key here")
res = client.embeddings.create(model="text-embedding-3-small", input="hello world")
print(len(res.data[0].embedding))

Limits and errors

The key’s requests per minute and tokens per minute limits apply, with the same x-ratelimit-* headers as chat completions. An unknown or non embedding model returns 404 model_not_found. With no credits left the call is refused with 402 insufficient_credits, unless your organization has its own key for that vendor. See Errors.