G2

Chetana M.

software engineer

Small-Business (50 or fewer emp.)

8/30/2026

Guest users or non-business users of the software, not included in G2 scores.

"Cohere Excels at Document Search with Strong Embedding and Reranking"

4.5/5

What do you like best about Cohere?

What I like most about Cohere is that it is pretty good when working with large amounts of text. The embed and Rerank models are useful for finding the information which is actually related to the query. I also like that I can use it for different AI workflows instead of only using it for normal chat. For RAG projects especially, the retrieval part is where I found it most useful. It works pretty well with the other tools in my stack, especially Langchain. I can use Cohere for embeddings, reranking, and LLM while Langchain handles the rest of the workflow. The setup takes a little time in the beginning, but once it is configured, moving the Cohere models into workflow is pretty smooth. Review collected by and hosted on G2.com.

What do you dislike about Cohere?

The main thing I don't like is choosing the right model can be a little confusing in the beginning. There are different models for chat, embeddings, and reranking, so it takes some time to understand which one to use. The API can also become expensive when you start using it more frequently. And sometimes I still need to change the prompt a few times to get the output I want. Review collected by and hosted on G2.com.

What problems is Cohere solving and how is that benefiting you?

Cohere helps me solve the problem of getting useful information from a large collection of text. I can use embeddings to find relevant content and then use reranking to put the better results on top. This is useful when building RAG applications because the model gets better context before generating the answer. It saves me from manually going through all the documents every time. Review collected by and hosted on G2.com.