# Inconsistent metadata filtering with serverless index

**URL:** <https://community.pinecone.io/t/inconsistent-metadata-filtering-with-serverless-index/7781>\
**Category:** Support\
**Tags:** vector-database, embeddings, serverless\
**Created:** [April 2, 2025, 1:31pm UTC](https://community.pinecone.io/t/inconsistent-metadata-filtering-with-serverless-index/7781 "2025-04-02T13:31:21Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![doriro](https://sea2.discourse-cdn.com/flex020/user_avatar/community.pinecone.io/doriro/32/3614_2.png) [@doriro](https://community.pinecone.io/u/doriro)\
**Post date:** [April 2, 2025, 1:31pm UTC](https://community.pinecone.io/t/inconsistent-metadata-filtering-with-serverless-index/7781/1 "2025-04-02T13:31:21Z")

</div>

We currently have a serverless index with the following index stats:

‘dimension’: 768,  
single namespace,  
‘total\_vector\_count’: 2096563

When running this query:

```auto
                    response = self.index.query(
                        vector=dummy_vector,
                        filter={'nms_id': {'$eq': '990049354070205171'}},
                        top_k=10000,
                        include_metadata=True
                    )

```

I’m getting 966 items as a result, all match the filter.  
If I try a different dummy value, I get 966 results, some of them are different.

Why does this inconsistency happen? This also happens in the pinecone ui.  
Thanks in advance!