# Hybrid Search using cosine similarity

**URL:** <https://community.pinecone.io/t/hybrid-search-using-cosine-similarity/6287>\
**Category:** Support\
**Tags:** semantic-search\
**Created:** [August 2, 2024, 12:39pm UTC](https://community.pinecone.io/t/hybrid-search-using-cosine-similarity/6287 "2024-08-02T12:39:06Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![thomas2](https://avatars.discourse-cdn.com/v4/letter/t/aeb1de/32.png) [@thomas2](https://community.pinecone.io/u/thomas2)\
**Post date:** [August 2, 2024, 12:39pm UTC](https://community.pinecone.io/t/hybrid-search-using-cosine-similarity/6287/1 "2024-08-02T12:39:06Z")

</div>

Hi,

I have implemented a similarity search RAG application using Pinecone’s hybrid search.

I want to implement a hybrid search with cosine similarity but I’m unsure if it can be done without customizations.

Previously, I implemented the same architecture with Langchain, which supported semantic search with Pinecone using cosine similarity in combination with the BM25 algorithm. However, I stopped using Langchain and now rely solely on Pinecone for semantic search.

Is it possible to achieve hybrid search with Pinecone using cosine similarity?

---

<div class="post-metadata">

**Author:** ![evan\_pinecone](https://sea2.discourse-cdn.com/flex020/user_avatar/community.pinecone.io/evan_pinecone/32/3023_2.png) [@evan\_pinecone](https://community.pinecone.io/u/evan_pinecone)\
**Post date:** [August 5, 2024, 7:32pm UTC](https://community.pinecone.io/t/hybrid-search-using-cosine-similarity/6287/2 "2024-08-05T19:32:44Z")

</div>

@thomas2 ,

With Pinecone, sparse-dense vectors can only be queried using the dot product metric. This is stated in [Limitations.](https://docs.pinecone.io/guides/data/understanding-hybrid-search#limitations)

Best,

Evan