Embeddings multilingual-e5-large pour un RAG en français

multilingual-e5-large pour un RAG en français : 1024 dimensions, 512 tokens, préfixes query:/passage: obligatoires, distance cosinus. Licence MIT.

Notes

Les préfixes « query: » et « passage: » sont obligatoires : sans eux, la recherche est nettement moins bonne.
512 tokens maximum par texte : découper les documents en morceaux d'environ 350 mots.
Normaliser les vecteurs et utiliser la distance cosinus.
Modèle publié par Microsoft (intfloat) sous licence MIT.

Estimated cost

Input tokens per call : ≈ 249

Input cost only, excluding the model's answer. Providers' public standard prices (Mistral AI, DeepSeek, OpenAI, Anthropic) converted from USD to euros at the ECB rate of September 29, 2026. Estimate: 1 token ≈ 3.6 characters.

Identifier

contextetech--multilingual-e5-large

Type

Embeddings

File

multilingual-e5-large.embedding.json

Format

JSON (application/json)

Size

686 bytes

Encoding

UTF-8

Estimated tokens

≈ 190

Language

French

License

MIT

Commercial use

Allowed

Personal data

None

Version

1.0

Published on

October 1, 2026 at 6:21 PM

Updated on

October 1, 2026 at 6:21 PM

Storage

in database

Uses

0

Likes

0

multilingual-e5-large.embedding.json

{
  "format": "contextetech/embedding/v1",
  "id": "contextetech/multilingual-e5-large",
  "description": "multilingual-e5-large pour un RAG en français : 1024 dimensions, 512 tokens, préfixes query:/passage: obligatoires, distance cosinus. Licence MIT.",
  "tags": [
    "rag",
    "multilingue",
    "recherche-semantique"
  ],
  "license": "MIT",
  "language": "fr",
  "model": "intfloat/multilingual-e5-large",
  "dimensions": 1024,
  "maxTokens": 512,
  "distance": "cosine",
  "languages": [
    "fr",
    "en",
    "de",
    "es",
    "it",
    "nl",
    "pt"
  ],
  "prefixes": {
    "query": "query: ",
    "passage": "passage: "
  },
  "chunkSize": 350,
  "benchmarks": []
}

Embed (Python)

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("intfloat/multilingual-e5-large")
q = model.encode(["query: your input here"], normalize_embeddings=True)
docs = model.encode(["passage: …"], normalize_embeddings=True)
print(q.shape)  # (1, 1024)
print(q @ docs.T)  # similarité