"AI Embeddings Comparison 2026: Choosing the Right Model for Your Application"

from openai import OpenAI

# Use Crazyrouter for lower prices
client = OpenAI(
    api_key="your-crazyrouter-key",
    base_url="https://api.crazyrouter.com/v1"
)

response = client.embeddings.create(
    model="text-embedding-3-small",
    input=["How to build a RAG application", "Retrieval augmented generation tutorial"],
    encoding_format="float"
)

embedding_1 = response.data[0].embedding  # 1536 dimensions
embedding_2 = response.data[1].embedding

# Calculate similarity
import numpy as np
similarity = np.dot(embedding_1, embedding_2) / (np.linalg.norm(embedding_1) * np.linalg.norm(embedding_2))
print(f"Similarity: {similarity:.4f}")

# With dimension reduction for storage efficiency
response = client.embeddings.create(
    model="text-embedding-3-large",
    input="Your text here",
    dimensions=1024  # Reduce from 3072 to 1024 with minimal quality loss
)

embedding = response.data[0].embedding  # Now 1024 dimensions

import cohere

co = cohere.Client("your-cohere-key")

# Cohere supports different input types for better results
response = co.embed(
    texts=["How to build a RAG application"],
    model="embed-v4",
    input_type="search_document",  # or "search_query", "classification", "clustering"
    embedding_types=["float"]
)

embedding = response.embeddings.float[0]

import voyageai

vo = voyageai.Client(api_key="your-voyage-key")

result = vo.embed(
    ["How to build a RAG application"],
    model="voyage-3-large",
    input_type="document"  # or "query"
)

embedding = result.embeddings[0]  # 2048 dimensions

# Via Crazyrouter
client = OpenAI(
    api_key="your-crazyrouter-key",
    base_url="https://api.crazyrouter.com/v1"
)

response = client.embeddings.create(
    model="text-embedding-005",
    input="How to build a RAG application"
)

embedding = response.data[0].embedding  # 768 dimensions

from sentence_transformers import SentenceTransformer

model = SentenceTransformer("BAAI/bge-m3")

texts = [
    "How to build a RAG application",
    "Retrieval augmented generation tutorial",
    "Best pizza recipe"
]

embeddings = model.encode(texts, normalize_embeddings=True)

# Calculate pairwise similarities
from sklearn.metrics.pairwise import cosine_similarity
similarities = cosine_similarity(embeddings)
print(similarities)

Tight budget ($0-10/month):
  → Google text-embedding-005 ($0.00625/M tokens)
  → Or self-host BGE-M3 (free, needs GPU)

Medium budget ($10-100/month):
  → text-embedding-3-small via Crazyrouter ($0.016/M tokens)
  → Best balance of cost and quality

High budget ($100+/month):
  → Voyage voyage-3-large ($0.18/M tokens)
  → Or text-embedding-3-large via Crazyrouter ($0.10/M tokens)

from openai import OpenAI
import numpy as np

client = OpenAI(
    api_key="your-crazyrouter-key",
    base_url="https://api.crazyrouter.com/v1"
)

# Step 1: Embed your documents
documents = [
    "Crazyrouter provides access to 300+ AI models through a single API.",
    "The pricing for GPT-5 through Crazyrouter is 20-30% cheaper than direct.",
    "Crazyrouter supports streaming, function calling, and vision models.",
    "You can use the OpenAI Python SDK with Crazyrouter by changing the base URL.",
]

doc_embeddings = []
for doc in documents:
    response = client.embeddings.create(
        model="text-embedding-3-small",
        input=doc
    )
    doc_embeddings.append(response.data[0].embedding)

doc_embeddings = np.array(doc_embeddings)

# Step 2: Embed the query
query = "How much does GPT-5 cost on Crazyrouter?"
query_response = client.embeddings.create(
    model="text-embedding-3-small",
    input=query
)
query_embedding = np.array(query_response.data[0].embedding)

# Step 3: Find most relevant documents
similarities = np.dot(doc_embeddings, query_embedding)
top_indices = np.argsort(similarities)[::-1][:2]  # Top 2

context = "\n".join([documents[i] for i in top_indices])

# Step 4: Generate answer with context
response = client.chat.completions.create(
    model="gpt-5",
    messages=[
        {"role": "system", "content": f"Answer based on this context:\n{context}"},
        {"role": "user", "content": query}
    ]
)

print(response.choices[0].message.content)