Embedding creator indeterminism

I’ve been using embedding creator using “text-similarity-babbage-001” and it seems that when using batch it returns different results on the input. When sending one by one for embeddings I always get the same result.

Is this something that is known? is there a workaround this?

for example sending this request 10 times:

curl --location --request POST 'https://api.openai.com/v1/embeddings' \
--header 'Content-Type: application/json' \
--header 'Accept: application/json' \
--header 'Authorization: Bearer XXXX \
--data-raw '{
"model": "text-similarity-babbage-001",
"input": ["Conduct legal research and gather evidence.", "Explain the law and give legal advice.", "Manage regulatory and compliance-related services."]
}' > 10.txt

9 times I get same result, in the 10th time I get a totally different result: