# The best embedding model for French: RAG on French documents

> French retrieval brings its own problems: missing accents, elisions, formal and informal registers, and administrative jargon. What to watch for, how to build a French test set, and a benchmark on French administrative and support documents.

Published 2026-10-06, updated 2026-11-14 by Truvec team. Canonical: https://truvec.dev/blog/best-embedding-model-for-french/

French is one of the best-supported languages after English, so most multilingual embedding models handle it reasonably well. "Reasonably well" isn't good enough for a support bot or a public-service assistant, though, and the differences between models on French are larger than their English scores suggest.

This post covers what makes French retrieval different, how to test it, and which models to compare. The last section benchmarks eight models on French administrative guides and support articles.

## What makes French different

**Accents, often missing.** Documents are written correctly ("délai", "résilier"), but users often type without accents ("delai", "resilier"), especially on mobile or on keyboards without French layouts. A good model treats both the same; a weaker one sees two different words.

**Elision and contraction.** "L'amende", "d'un", "qu'il", "au", "du": articles merge with the next word. Tokenizers split these in different ways, and models that saw little French may handle them poorly.

**Register.** Documents are formal ("Vous disposez d'un délai de 45 jours pour contester"), questions are often informal ("j'ai combien de temps pour contester une amende ?"). The model has to match the two.

**Acronyms and jargon.** French administration and business run on acronyms (CAF, URSSAF, APL, SIRET, TVA) that general training data barely covers, and on their own phrasing ("justificatif de domicile", "avis d'imposition").

**Anglicisms and English terms.** Tech and business documents mix in English words ("le dashboard", "un webhook", "le cloud"), sometimes alongside French equivalents ("tableau de bord").

**Several Frenches.** European French, Canadian French and African varieties differ in vocabulary ("courriel" vs "e-mail", "magasiner" vs "faire du shopping"). If your users are in Quebec, test with Quebec phrasing.

## Build a French test set

- Generate questions **in French**, not translated from English: translations sound unnatural and favor models that are good at translated text.
- Include questions **without accents** and in an **informal register**, the way users type.
- Add questions that use **acronyms** and others that spell out the full term.
- If your corpus mixes French and English, add questions in each language.

## Which models to test

- **Multilingual API models**: Cohere Embed v4 and Embed Multilingual v3, Gemini Embedding, OpenAI text-embedding-3-large and 3-small.
- **Open-weight**: Qwen3 Embedding, trained on many languages and self-hostable.
- **An English-only model as a baseline** (such as BGE Base EN): it shows the gap a multilingual model closes.

French data is often personal or sensitive (administrative files, customer records). If GDPR or a client contract restricts where the data goes, check where each provider processes it, or consider a self-hosted open model.

The benchmark below runs eight models on French administrative guides and support articles. Look at the example questions, which are written the way French users actually ask. For what each metric in the tables means, see [RAG evaluation metrics explained](/blog/rag-evaluation-metrics-that-matter/); for why a public leaderboard can't settle this for your documents, see [why MTEB alone isn't enough](/blog/mteb-benchmark-validate-embedding-model-before-switching/).

## Benchmark: Embedding models on French administrative guides and help pages

Real benchmark generated by Truvec (benchmark #13, run 2026-11-14). Raw data: [benchmark.json](https://truvec.dev/blog/best-embedding-model-for-french/benchmark.json), [benchmark.csv](https://truvec.dev/blog/best-embedding-model-for-french/benchmark.csv). Full report with every question: [report.pdf](https://truvec.dev/blog/best-embedding-model-for-french/report.pdf).

- Corpus: service-public.fr practical guides and French Wikipedia help pages (https://www.service-public.gouv.fr), 300 documents, 4,345 chunks
- Chunking: recursive, 1000 chars, 200 overlap
- Golden set: 100 questions generated by gpt-5.6-luna, human-reviewed
- Retrieval: top 5 by cosine similarity; models ranked by MRR, then recall@k

### Key results

- **Qwen3 Embedding 8B** ranks first, with an MRR of 0.909 and a hit rate of 98.0% in the top 5.
- One model is within 0.02 MRR of the winner (Cohere Embed v4): on this corpus they are practically tied.
- Hit rate ranges from 75.0% to 100.0% across the 8 models.
- Cheapest model within 2 points of the winner's hit rate: **OpenAI text-embedding-3-small**, $0.02 to embed this corpus against $0.10.
- Fastest to embed a query: **OpenAI text-embedding-3-small** (7.1 ms per query, measured in batch from our servers).

### Retrieval quality

| Rank | Model | Provider | Hit rate@5 | Recall@5 | MRR | nDCG@5 | Precision@5 |
| --- | --- | --- | ---: | ---: | ---: | ---: | ---: |
| 1 | Qwen3 Embedding 8B | fireworks | 98.0% | 98.0% | 0.909 | 0.927 | 19.6% |
| 2 | Cohere Embed v4 | cohere | 99.0% | 99.0% | 0.890 | 0.915 | 19.8% |
| 3 | Cohere Embed Multilingual v3 | cohere | 99.0% | 99.0% | 0.875 | 0.904 | 19.8% |
| 4 | Voyage 4 Large | voyage | 97.0% | 97.0% | 0.865 | 0.891 | 19.4% |
| 5 | OpenAI text-embedding-3-large | openai | 100.0% | 100.0% | 0.847 | 0.885 | 20.0% |
| 6 | Gemini Embedding 2 | google | 99.0% | 99.0% | 0.837 | 0.875 | 19.8% |
| 7 | OpenAI text-embedding-3-small | openai | 97.0% | 97.0% | 0.818 | 0.856 | 19.4% |
| 8 | BGE Base EN v1.5 | fireworks | 75.0% | 75.0% | 0.612 | 0.647 | 15.0% |

### Hit rate by position

| Model | Top 1 | Top 2 | Top 3 | Top 4 | Top 5 |
| --- | ---: | ---: | ---: | ---: | ---: |
| Qwen3 Embedding 8B | 86.0% | 93.0% | 96.0% | 96.0% | 98.0% |
| Cohere Embed v4 | 82.0% | 92.0% | 96.0% | 97.0% | 99.0% |
| Cohere Embed Multilingual v3 | 79.0% | 91.0% | 97.0% | 98.0% | 99.0% |
| Voyage 4 Large | 79.0% | 90.0% | 94.0% | 95.0% | 97.0% |
| OpenAI text-embedding-3-large | 74.0% | 90.0% | 94.0% | 97.0% | 100.0% |
| Gemini Embedding 2 | 73.0% | 89.0% | 94.0% | 94.0% | 99.0% |
| OpenAI text-embedding-3-small | 73.0% | 82.0% | 91.0% | 94.0% | 97.0% |
| BGE Base EN v1.5 | 50.0% | 68.0% | 73.0% | 75.0% | 75.0% |

### Cost, speed and size

| Model | Price / 1M tokens | Embed the corpus | 1,000 queries | Query embedding | Search p95 | Vectors |
| --- | ---: | ---: | ---: | ---: | ---: | ---: |
| Qwen3 Embedding 8B | $0.10 | $0.10 | $0.0031 | 10 ms | 60 ms | 67.9 MB |
| Cohere Embed v4 | $0.12 | $0.11 | $0.0037 | 20 ms | 69 ms | 25.5 MB |
| Cohere Embed Multilingual v3 | $0.10 | $0.10 | $0.0031 | 8.8 ms | 52 ms | 17.0 MB |
| Voyage 4 Large | $0.12 | $0.11 | $0.0037 | 15 ms | 58 ms | 17.0 MB |
| OpenAI text-embedding-3-large | $0.13 | $0.12 | $0.0040 | 9.0 ms | 90 ms | 50.9 MB |
| Gemini Embedding 2 | $0.20 | $0.19 | $0.0062 | 20 ms | 59 ms | 50.9 MB |
| OpenAI text-embedding-3-small | $0.02 | $0.02 | $0.0006 | 7.1 ms | 45 ms | 25.5 MB |
| BGE Base EN v1.5 | $0.0080 | $0.0077 | $0.0002 | 19 ms | 51 ms | 12.7 MB |

### Score separation

| Model | Relevant chunk | Best distractor | Gap |
| --- | ---: | ---: | ---: |
| Qwen3 Embedding 8B | 0.829 | 0.780 | 0.050 |
| Cohere Embed v4 | 0.672 | 0.616 | 0.057 |
| Cohere Embed Multilingual v3 | 0.796 | 0.756 | 0.041 |
| Voyage 4 Large | 0.637 | 0.596 | 0.042 |
| OpenAI text-embedding-3-large | 0.729 | 0.689 | 0.040 |
| Gemini Embedding 2 | 0.824 | 0.798 | 0.025 |
| OpenAI text-embedding-3-small | 0.770 | 0.736 | 0.033 |
| BGE Base EN v1.5 | 0.763 | 0.743 | 0.019 |

### Example questions

Rank of the first relevant chunk per model ("miss": not in the top 5).

| Question | Qwen3 Embedding 8B | Cohere Embed v4 | Cohere Embed Multilingual v3 | Voyage 4 Large | OpenAI text-embedding-3-large | Gemini Embedding 2 | OpenAI text-embedding-3-small | BGE Base EN v1.5 |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| Après la validation de la demande en ligne, quel délai faut-il généralement prévoir pour recevoir son permis de conduire ? | 1 | miss | 1 | miss | 2 | 5 | 4 | 3 |
| Quelles sont les conditions pour utiliser un sas Parafe à l’arrivée en France ? | 5 | 4 | 2 | miss | 1 | 5 | 3 | 1 |
| Quels territoires sont concernés par une interdiction de retour lorsqu’une IRTF est prononcée en France ? | 5 | 5 | 3 | 2 | 5 | 5 | miss | 4 |
| Que dois-je faire auprès de ma caisse de Sécurité sociale si ma situation professionnelle change, par exemple si je pars à la retraite ou change d'activité ? | 1 | 5 | 4 | 4 | 4 | 3 | 3 | miss |
| Où une victime peut-elle déposer plainte, et quels documents peut-elle obtenir à la fin de son audition ? | 3 | 2 | 2 | 3 | 2 | 5 | miss | miss |
| Comment faire enregistrer une modification de convention de Pacs et quelles pièces faut-il présenter en mairie ? | 1 | 1 | 2 | 2 | 3 | 3 | 2 | miss |
| Comment un locataire peut-il contester un complément de loyer, et dans quel délai ? | 2 | 2 | 3 | 3 | 1 | 2 | 1 | miss |
| Dans quelles situations un véhicule d’occasion peut-il circuler avec un certificat W garage ? | 2 | 1 | 3 | 1 | 1 | 3 | 3 | miss |

### Notes

- The corpus is 230 practical guides (fiches pratiques) for individuals from the service-public.fr open data, published under Licence Ouverte 2.0, and 70 help pages from French Wikipedia (Aide namespace, CC BY-SA 4.0), fetched on 14 November 2026. Guides were picked at random among those between 5,000 and 30,000 characters, spread across themes (justice, family, housing, money and taxes, work, transport, health, foreigners, papers and citizenship).
- All 100 questions are written in French. 98 are about the administrative guides and 2 about the help pages: questions are drawn from chunks at random, and the guides make up most of the text.
- BGE Base EN v1.5 is English-only: it is included as a baseline.
- With 100 questions, a gap of a couple of points between two models is within the noise.

### Metric definitions

- **Hit rate**: Share of questions with at least one relevant chunk in the top k results. The most direct answer to "does the right passage reach the LLM?".
- **Recall**: Share of the relevant chunks found in the top k, averaged over questions. Matters when an answer spans several chunks.
- **MRR**: Mean reciprocal rank: 1 when the first relevant chunk is ranked first, 1/2 when second, 0 when missing. Rewards putting the answer at the top.
- **nDCG**: Normalized discounted cumulative gain: how close the ranking of all relevant chunks is to the ideal ranking.
- **Precision**: Share of the top k results that are relevant. Low precision means more noise sent to the LLM.
