Embedding Model Comparison: OpenAI vs Cohere vs Open Source

Embedding Model Comparison: OpenAI vs Cohere vs Open Source

💡 What You Will Learn

Embedding Model Comparison: OpenAI vs Cohere vs Open Source

from sentence_transformers import SentenceTransformer
import numpy as np

query = ""
docs = ["30", "iPhone15", "200"]

model = SentenceTransformer("BAAI/bge-large-zh-v1.5")
query_vec = model.encode(query)
doc_vecs = model.encode(docs)

scores = np.dot(doc_vecs, query_vec) / (
    np.linalg.norm(doc_vecs, axis=1) * np.linalg.norm(query_vec)
)
for doc, score in zip(docs, scores):
    print(f"{score:.3f} -> {doc}")
# 0.892 -> ...
# 0.451 -> 
# 0.523 -> 

|:----|:---:|:----:|:-------:| | text-embedding-3-small | 1536 || 62.1 | | text-embedding-3-large | 3072 || 64.5 | | BAAI/bge-large-zh-v1.5 | 1024 || 65.8 | | shibing624/text2vec | 768 || 63.2 | | Cohere embed-multilingual-v3 | 1024 || 63.9 |

Recommendations