Embedding

Embedding models convert text into dense vectors that represent semantic information, making semantically similar texts closer in vector space. They are commonly used for semantic search, retrieval, reranking, clustering, and RAG (Retrieval-Augmented Generation), and are served through an OpenAI-compatible Embeddings API. Representative models include:

Qwen/Qwen3-Embedding-8B
Qwen/Qwen3-Embedding-4B
Qwen/Qwen3-Embedding-0.6B
IEITYuan/Yuan-embedding-2.0-en

For the full model list and pricing, refer to Models.

Core Capabilities

Semantic search: Match queries and documents by meaning rather than exact keywords.
Retrieval for RAG: Embed knowledge base content and store the vectors, then retrieve the most relevant chunks to provide context for LLM answers.
Clustering and deduplication: Group similar items or detect near-duplicate content.
Recommendation and reranking: Score candidates by similarity to a query or user profile.

Usage

The Embeddings API accepts a single string or a list of strings and returns a vector for each input. Common workflow:

Generate vectors for documents and store them.
When a user sends a query, embed the query text.
Compare the query vector with existing vectors, for example by using cosine similarity.

Key Parameters

model: The Embedding model name.
input: A string or a list of strings.
encoding_format: Accepted by some models, for example an encoding_format of float.

Notes

Use the same Embedding model for documents and queries so their vectors are in the same space and can be compared.
If an input exceeds the model’s maximum length, split the long text into smaller chunks before embedding. The maximum input length (context window), pricing, and supported languages vary by model and are shown in the model details in Models.
When embedding large amounts of text, batch multiple inputs into a single request to reduce round trips.

Billing

Formula: Total cost = input tokens x input unit price. Embedding model requests are billed only by input tokens and do not produce generated output.
Pricing: Check the model detail page in Models for each model’s pricing.
Recommendations:
- Documents usually only need to be embedded once, and the results should be cached to avoid repeated computation on every request.
- Prefer batch processing to improve request efficiency.

Examples

The examples read the API key from an environment variable to avoid writing secrets into code.

export API_KEY="YOUR_API_KEY"

Embed Text (Python)

import os
import requests

api_key = os.environ["API_KEY"]
api_url = "https://api.scitix.ai/model-api/embeddings"

response = requests.post(
    api_url,
    headers={
        "Authorization": f"Bearer {api_key}",
        "Content-Type": "application/json",
    },
    json={
        "model": "Qwen/Qwen3-Embedding-8B",
        "input": [
            "Machine learning is a subset of artificial intelligence",
            "Deep learning uses neural networks with multiple layers",
            "Natural language processing enables computers to understand text",
        ],
    },
)

data = response.json()
for i, item in enumerate(data["data"]):
    print(f"Embedding {i}: {len(item['embedding'])} dimensions")

Embed Text (curl)

curl -X POST https://api.scitix.ai/model-api/embeddings \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $API_KEY" \
  -d '{
    "model": "Qwen/Qwen3-Embedding-8B",
    "input": [
      "Machine learning is a subset of artificial intelligence",
      "Deep learning uses neural networks with multiple layers"
    ]
  }'

Semantic Similarity

import os
import requests
import numpy as np

api_key = os.environ["API_KEY"]
api_url = "https://api.scitix.ai/model-api/embeddings"


def embed(texts):
    resp = requests.post(
        api_url,
        headers={
            "Authorization": f"Bearer {api_key}",
            "Content-Type": "application/json",
        },
        json={"model": "Qwen/Qwen3-Embedding-8B", "input": texts},
    )
    return [np.array(item["embedding"]) for item in resp.json()["data"]]


def cosine(a, b):
    return float(a @ b / (np.linalg.norm(a) * np.linalg.norm(b)))


query, *docs = embed([
    "How do I reset my password?",
    "Steps to recover a forgotten account password.",
    "Our refund policy for annual plans.",
])

for i, doc in enumerate(docs):
    print(f"doc {i}: {cosine(query, doc):.3f}")

Get Started

Models

Usage

API Guides

Tool Integrations

Core Capabilities

Usage

Key Parameters

Notes

Billing

Examples

Embed Text (Python)

Embed Text (curl)

Semantic Similarity

​Core Capabilities

​Usage

​Key Parameters

​Notes

​Billing

​Examples

​Embed Text (Python)

​Embed Text (curl)

​Semantic Similarity

Core Capabilities

Usage

Key Parameters

Notes

Billing

Examples

Embed Text (Python)

Embed Text (curl)

Semantic Similarity