Model reference · open weights
Solon-embeddings-large-0.1 is an open-weight embedding model from OrdalieTech. Solon-embeddings-large-0.1 (FP32) weighs 1.1 GB; the smallest configuration that runs it is RTX 3060 12 GB.
Summary of the OrdalieTech/Solon-embeddings-large-0.1 model card, 2026-10-01
What it is
| Released by | OrdalieTech |
|---|---|
| Released | 2023-12-09 |
| Parameters | 560M |
| VRAM | 1.1 GB for the weights |
What it runs on
| Card | Runs | Memory |
|---|---|---|
| RTX 3060 12 GB | fits | 11.6 GB |
| RTX 4060 Ti 16 GB | fits | 15.4 GB |
| RTX 3090 24 GB | fits | 23.4 GB |
| RTX 4090 24 GB | fits | 23.4 GB |
| RTX 5090 32 GB | fits | 31.0 GB |
| L40S 48 GB | fits | 44.0 GB |
| A100 80 GB | fits | 78.2 GB |
| H100 80 GB | fits | 78.1 GB |
| RTX PRO 6000 Blackwell 96 GB | fits | 93.8 GB |
| DGX Spark (GB10) 128 GB unified | fits | 107 GB |
| H200 141 GB | fits | 138 GB |
| B200 180 GB | fits | 176 GB |
From the model card
SOTA Open source french embedding model.
Instructions : Add "query : " before the query to retrieve to increase performance of retrieval. No instructions needed for passages.
| Model | Mean Score |
|---|---|
| OrdalieTech/Solon-embeddings-large-0.1 | 0.7490 |
| cohere/embed-multilingual-v3 | 0.7402 |
| OrdalieTech/Solon-embeddings-base-0.1 | 0.7306 |
| openai/ada-002 | 0.7290 |
| cohere/embed-multilingual-light-v3 | 0.6945 |
| antoinelouis/biencoder-camembert-base-mmarcoFR | 0.6826 |
| dangvantuan/sentence-camembert-large | 0.6756 |
| voyage/voyage-01 | 0.6753 |
| intfloat/multilingual-e5-large | 0.6660 |
| intfloat/multilingual-e5-base | 0.6597 |
| Sbert/paraphrase-multilingual-mpnet-base-v2 | 0.5975 |
| dangvantuan/sentence-camembert-base | 0.5456 |
| EuropeanParliament/eubert_embedding_v1 | 0.5063 |
These results have been obtained through 9 french benchmarks on a variety of text similarity tasks (classification, reranking, STS) :
We created OrdalieFRSTS and OrdalieFRReranking to enhance the benchmarking capabilities of French STS and reranking assessments.
(evaluation script available here : github.com/OrdalieTech/mteb)
Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.
Benchmarks
As published on the model card: the maker's own numbers, not measured by AxForge.
| Task | Dataset | Metric | Score |
|---|---|---|---|
| Passage Retrieval | mMARCO-fr | Recall@500 | 92.700 |
| Passage Retrieval | mMARCO-fr | Recall@100 | 82.700 |
| Passage Retrieval | mMARCO-fr | Recall@10 | 55.500 |
| Passage Retrieval | mMARCO-fr | MAP@10 | 29.400 |
| Passage Retrieval | mMARCO-fr | nDCG@10 | 35.800 |
| Passage Retrieval | mMARCO-fr | MRR@10 | 29.900 |
| Clustering | MTEB AlloProfClusteringP2P | v_measure | 64.169 |
| Clustering | MTEB AlloProfClusteringS2S | v_measure | 38.171 |
| Reranking | MTEB AlloprofReranking | map | 64.877 |
| Reranking | MTEB AlloprofReranking | mrr | 66.391 |
| Retrieval | MTEB AlloprofRetrieval | map_at_1 | 29.620 |
| Retrieval | MTEB AlloprofRetrieval | map_at_10 | 40.963 |
| Retrieval | MTEB AlloprofRetrieval | map_at_100 | 41.894 |
| Retrieval | MTEB AlloprofRetrieval | map_at_1000 | 41.939 |
| Retrieval | MTEB AlloprofRetrieval | map_at_3 | 37.709 |
| Retrieval | MTEB AlloprofRetrieval | map_at_5 | 39.697 |
| Retrieval | MTEB AlloprofRetrieval | mrr_at_1 | 29.620 |
| Retrieval | MTEB AlloprofRetrieval | mrr_at_10 | 40.963 |
| Retrieval | MTEB AlloprofRetrieval | mrr_at_100 | 41.894 |
| Retrieval | MTEB AlloprofRetrieval | mrr_at_1000 | 41.939 |
| Retrieval | MTEB AlloprofRetrieval | mrr_at_3 | 37.709 |
| Retrieval | MTEB AlloprofRetrieval | mrr_at_5 | 39.697 |
| Retrieval | MTEB AlloprofRetrieval | ndcg_at_1 | 29.620 |
| Retrieval | MTEB AlloprofRetrieval | ndcg_at_10 | 46.942 |