intfloat/multilingual-e5-large

multilingual‑e5‑large is a Microsoft Research text embedding model based on XLM‑RoBERTa‑large and trained on multilingual datasets. With 24 layers and 1024-dimensional embeddings, it supports 100 languages for semantic search, text similarity, bitext mining and classification. It was weakly supervised pre-trained and then fine-tuned on datasets such as MS MARCO and MIRACL. Inputs are truncated to 512 tokens.

• PROVIDER LOCATION: Extra UE
• PARAMETERS: ~560M
• MAX CONTEXT: 512 token
• LICENSE: MIT
• FEATURES: Embedding retrieval, Semantic similarity
• OUTPUT FORMATS: Embedding_vector

• Multilingual semantic search
• RAG systems (Retrieval Augmented Generation)
• Text classification
• Document clustering and deduplication

curl --request POST \
--url https://apiaruba.it/aiplatform/v1/embeddings \
--header 'content-type: application/json' \
--header 'x-aruba-api-key: *****' \
--data '{
"model": "multilingual-e5-large",
"input": [
"primo testo",
"secondo testo"
]
}'

{
"model": "multilingual-e5-large",
"data": [
{
"embedding": [
0.02086777612566948,
0.003188132541254163,
0.011166091077029705,
-0.066386379301548,
0.04100335016846657,
...
]
}
]
}

• BGE‑M3

• Hugging Face
• Paper
• Azure AI Catalog