Uncover trends and patterns in text

Turn text into numerical representations of language for deeper insights at scale. Embed makes it possible to algorithmically categorize and score text quickly to extract meaning.

Extract insights with an API call

No matter your level of experience with ML/AI, the Cohere Platform makes it easy to integrate language comprehension into your application.

Multiple platforms supported

We support all common languages through native SDK support or encapsulated REST calls, see examples

Large language models

Our models have been trained on billions of words, allowing them to understand nuance and context.

import cohere
co = cohere.Client('{apiKey}')
response = co.embed(
texts=["When are you open?", "When do you close?", "What are the hours?", "Do you have a vegan option?", "Do you have vegetarian?"])
print('Embeddings: {}'.format(response.embeddings))

How It Works

Embeddings are numerical representations of meaning in text. Because they are numbers, they can be compared to each other for similarity. They can also be plotted on a chart that shows which texts are similar to each other. Large language models produce highly nuanced embeddings.

Why Cohere

Free Developer Tier

Learn and iterate with the Cohere API free of charge until you go to production.

Easy To Use

No prior ML/AI experience required. Get started with just a few examples.

Customizable Models

Customize our models with your own data sets, and deploy them easily to production.

What's possible with Embed

Semantic Search

Enable users to search using conversational language.

Topic Modeling

Cluster similar topics and discover thematic trends across a body of text sources.


Build a recommendation engine and engage your users with more relevant content.

Multilingual embeddings

Run topic modeling, semantic search, recommendations across 100+ languages with just one model. Read more here.

Totally transparent pricing

We're making it easy to explore, learn, and experiment with the Cohere Platform.


All models


Limit of 100 calls/minute


Default models


per 1000 Embeddings

Custom models


per 1000 Embeddings


Using a default model, embedding 10,000 survey responses to cluster by themes would cost $10
Embedding 1,000 podcast titles with a custom model to improve recommendation relevance would cost $2