MTEB Benchmark Results (2025)
| Model | Avg Score | Dim | Cost |
|---|---|---|---|
| text-embedding-3-large | 64.6 | 3072 | $0.13/M tokens |
| Cohere embed-v3 | 64.5 | 1024 | $0.10/M tokens |
| BGE-M3 | 63.8 | 1024 | Free |
| E5-mistral-7b | 66.6 | 4096 | Free |
| Jina-embeddings-v3 | 65.2 | 1024 | Free |
For Multilingual (AR/FR/EN)
BGE-M3 and Jina-v3 have the best multilingual coverage at zero cost.
My Stack
# Free, local, multilingual
from sentence_transformers import SentenceTransformer
model = SentenceTransformer('BAAI/bge-m3')
embeddings = model.encode(texts, normalize_embeddings=True)