How to beginner · 3 min read

How to set embed model in LlamaIndex

Q: How to set embed model in LlamaIndex

In LlamaIndex, set the embedding model by passing a custom embedding object to the ServiceContext or directly to the index constructor. Use embedding classes like OpenAIEmbedding with your preferred model name to specify the embed model.

Quick answer

In LlamaIndex, set the embedding model by passing a custom embedding object to the ServiceContext or directly to the index constructor. Use embedding classes like OpenAIEmbedding with your preferred model name to specify the embed model.

PREREQUISITES

Python 3.8+
OpenAI API key or other embedding API key
pip install llama-index openai

Setup

Install llama-index and openai packages, and set your API key as an environment variable.

Install packages: pip install llama-index openai
Set environment variable: export OPENAI_API_KEY='your_api_key' (Linux/macOS) or setx OPENAI_API_KEY "your_api_key" (Windows)

bash

pip install llama-index openai

Step by step

Use the OpenAIEmbedding class from llama_index.embeddings.openai to specify the embedding model, then pass it to ServiceContext and create the index.

python

import os
from llama_index import SimpleDirectoryReader, ServiceContext, GPTVectorStoreIndex
from llama_index.embeddings.openai import OpenAIEmbedding

# Set your OpenAI API key in environment variable
api_key = os.environ["OPENAI_API_KEY"]

# Create embedding instance with desired model
embedding = OpenAIEmbedding(model_name="text-embedding-3-large", openai_api_key=api_key)

# Create service context with custom embedding
service_context = ServiceContext.from_defaults(embed_model=embedding)

# Load documents
documents = SimpleDirectoryReader("./data").load_data()

# Create index with custom embedding via service context
index = GPTVectorStoreIndex.from_documents(documents, service_context=service_context)

# Query example
query_engine = index.as_query_engine()
response = query_engine.query("What is LlamaIndex?")
print(response.response)

output

LlamaIndex is a data framework for building LLM applications by connecting your data with large language models.

Common variations

Use other embedding providers by implementing the BaseEmbedding interface.
Pass the embedding directly to index constructors that accept an embed_model parameter.
Use async methods if supported by your embedding provider.

python

from llama_index import GPTVectorStoreIndex
from llama_index.embeddings.openai import OpenAIEmbedding

embedding = OpenAIEmbedding(model_name="text-embedding-3-large", openai_api_key=os.environ["OPENAI_API_KEY"])

# Directly pass embed_model to index
index = GPTVectorStoreIndex.from_documents(documents, embed_model=embedding)

Troubleshooting

If you get authentication errors, verify your API key is correctly set in os.environ.
If embeddings are not generated, check the model name is valid and supported by your embedding provider.
For slow responses, ensure network connectivity and consider using smaller embedding models.

✅

Key Takeaways

Use OpenAIEmbedding or a compatible embedding class to specify your embed model in LlamaIndex.
Pass the embedding instance to ServiceContext.from_defaults(embed_model=...) for centralized configuration.
You can also pass the embed model directly to index constructors that accept an embed_model parameter.
Always set your API keys securely via environment variables to avoid authentication issues.
Check model names against your embedding provider's documentation to ensure compatibility.

Verified 2026-04 · text-embedding-3-large

Verify ↗