Embeddings and reranking

Vectors for semantic search and reranking for RAG

Embedding APIs turn text, code or images into vectors, so your own search finds results by meaning instead of exact words; reranking APIs sort a list of search results by how well each one answers a question. Together they are the retrieval part of retrieval-augmented generation (RAG). Check the free token allowance and whether your vectors stay compatible when a model is updated.

Compare the embedding and reranking APIs

ServiceAPI keyFormatsFree use
Pinecone API Needed, free JSON Free on the Starter plan: up to 2 GB storage, 1 million read units and 2 million write units a month, 5 million embedding tokens a month per hosted model and 500 rerank requests a month.
Qdrant Cloud Needed, free JSON Free forever: one cluster with 1 GB RAM and 4 GB disk and free small embedding models, no credit card needed; unused free clusters are suspended after a week and deleted after four weeks.
Upstash Vector Needed, free JSON Free plan with up to 10 indexes, 1 GB of data and 10,000 queries and updates a day; pay as you go after adding a card.
Voyage AI API Needed, free JSON Free with a free API key for the first 200 million tokens per account on most current models; paid per token after that.

Questions

Are these embedding and reranking APIs really free?

The services here can be used without paying, at least on a free plan or free tier, often with limits such as a number of requests per day or non-commercial use only. The table above quotes what each service says about its free use; check the provider's own terms before you rely on it.

How up to date is this list?

The address of every service is checked about once a week, and services that stop working or stop being free are removed. This list last changed on November 21st, 2026.