Cohere Rerank 2 Model - English

My main use case for Cohere is to use a Cohere embedded model to create our own vector databases and check conversations.

A specific example of how I use Cohere's embedding model for our vector databases or conversation checking involves abilities that take customer approvals and convert that information into vectors. I save this information in our own systems and also store small vectors on customer devices to use during custom customer requests.

My use case involves indexing and saving small portions of information.

In my experience, Cohere offers reliable embedding models for customers who do not want to use standard OpenAI models.

I find that the choice of embedding models is limited, and Cohere was available for Azure , which makes it a good alternative for customers who prefer not to use OpenAI.

Cohere has positively impacted my organization by helping our customers work more efficiently when creating requests, and the embedding results are of very high quality.

I believe Cohere can be improved technically by providing more feedback, logs, and metrics for embedding requests, as it currently appears to be a black box without any understanding of quality. Quality can only be understood after using it with customer requests, and during the embedding process, measurable metrics are not visible.

There are no particularly unique features distinguishing Cohere from other solutions.

I have been using Cohere for approximately nine to ten months.

Cohere is stable in my experience.

The scalability of Cohere showed that after sending a large amount of information and embeddings, it became slower, though we do not use any special solution for scaling.

I have not interacted with Cohere's support team. However, I contacted Azure  about the slowness, and we decided to use smaller chunks of information during the embedding process.

I previously used embedding models from OpenAI. I switched to Cohere because customers wanted to use something other than OpenAI models.

I did not purchase Cohere through the Azure Marketplace . I deployed unmanaged models and shared models.

I do not have relevant metrics about the return on investment from using Cohere yet because the customer's application is in a paging stage and has not been released. However, I understand that it is performing well, and we plan to continue with it.

My experience with pricing, setup cost, and licensing indicates that it does not require a special license, and the prices are competitive compared to competitors.

I did not evaluate other options before choosing Cohere. I looked at prices, and since we used Azure cloud, it did not provide many models for selection. Only OpenAI and Cohere were available for embedding.

For others looking into using Cohere, I advise that it is a good model for people who want to be agnostic when using models and creating something flexible to switch from one model to another. I would rate this product an eight out of ten.

We founded this company two and a half years ago, and since the middle of 2022, we foresaw the trending of generative AI and large language models, so my startup is working on developing generative AI applications for our clients, including enterprises and a few other startups across America and Canada.

I started using Cohere when we first got information from the community about their reranking models almost one and a half years ago.

In some clients' projects, we were required to introduce reranking model in the RAG flow (Retrieval-augmented generation). In this flow, we use different components to allow users to select and pick up from the UI components, drag and drop to their flow to enhance their RAG pipeline. That's where we introduced Cohere models as one of the providers for reranking.

Cohere's reranking model helped us complete this request

From our data, I can tell that at least 15% of end users were actively using reranking to enhance their RAG pipeline because we have the UI to indicate that reranking is recommended as it can enhance the quality of the retrieval.

For clarification, I want to describe this data more clearly. As mentioned, 15% of end users chose to enable this module based on the fact that we have the pricing tier with an extra cost for their API call.

In general, I'm satisfied with the speed, and I can confirm this because we have the long fields to track all conversations, and we see that this loop for reranking actually costs relatively less time throughout the whole chat flow. Regarding quality, it's hard to tell because we don't have a benchmark. In our enterprise applications, we are trying to build up evaluation pipelines, do AB testing, and other analysis, but it's not a conventional computer science application, so it's very hard to build up evaluation pipelines with objective criteria. It's challenging for us to make a conclusion about quality, but the speed is good.

A direct benefit of using Cohere's reranking model is that we can tell clients we have this module rather than missing this piece, as reranking is a very important component that companies discuss to enhance RAG quality.

Although it's not impacting our business model, I'm pushing for the evaluation system because it can expand our business scope. We want to sell our system to clients, and while they may not be aware of evaluation initially, it's beneficial to have. Once we have these systems, we can showcase to end users that employing such a reranking system improves quality. We need proof to convince ourselves that after implementing reranking, we get better quality.

It would be better to have a dashboard for users to showcase how reranking helps improve quality. When end users choose the service, they want to see the actual output. The evaluation part is challenging for recent large language model applications but remains very important.

If Cohere could provide a dashboard where we can employ an LLM as a judge to check quality before and after reranking, that would be helpful. We could either have another large language model evaluate this part or allow UAT users to manually check with humans in the middle. As an enterprise provider, we want such features because when chatting with clients, we can demonstrate that employing Cohere's reranking model significantly improves results compared to not using it.

Documentation is not a major blocking issue for us as we are sophisticated software engineers. Integration and the API provided for reranking models are not complicated, so we can easily handle that. The documentation is good. The major point is to prove the value through evaluation. We need a sophisticated solution to showcase visibly to our clients and engineering team to convince them that using this model creates improvements.

I started using Cohere when we first got information from the community about their reranking models almost one and a half years ago.

That's only what we need in our product currently. I will communicate when we have other requirements.

We haven't had any issues to escalate to Cohere's support because reranking is an optional feature in our product, and we haven't seen any significant issues so far.

We don't observe many scaling problems because it's an enterprise application. There are a few hundred people using this. The concurrent user rate is not significant, which might be why we don't see many scaling issues so far.

We haven't had any issues to escalate to Cohere's support because reranking is an optional feature in our product, and we haven't seen any significant issues so far.

For reranking, Cohere was our only solution.

I'm more focused on the speed and overall quality of the model itself and the chat flow as a whole solution. That's why I'm not in the position to comment on the price and setup cost as there are DevOps working on this piece.

Hard to estimate the overall ROI. but if you see the ROI for the feature of reranking, it's a positive number

I'm not in the position to answer that question because I was not the one who deployed that model, but I believe it is because we see the model name as ARN name, so it's most likely coming from Bedrock.

For reranking, Cohere is the only solution we have used so far.

As a feature developer, I'm more focused on the speed and overall quality of the model itself and the chat flow as a whole solution. That's why I'm not in the position to comment on the price and setup cost as there are DevOps working on this piece. My rating for this solution is 8 out of 10.

I am working on test automation, specifically an intelligent test automation framework. Based on the existing framework, which is handled in TypeScript and Selenium, I used Cohere intelligence to create new tests based on the test data and test cases that we provide. It will read through all the test cases in natural language, process them, analyze the internal working of our existing framework, and create the artifacts, test data, and test source based on the existing framework.

Currently, we are using Cohere APIs. First, I used the chat in the application itself to identify how it works by providing RAG sources, including PDF and text files. After confirming it worked fine, we moved to find an API, and we are using that API to handle all these tasks. The APIs are very functional for all our current use cases, mainly the intelligent test automation.

Cohere is very useful because I have been in scenarios where code was written with multiple reusable concepts containing many functionalities covered as different functions, but without descriptions of what particular functions were doing. We used Cohere intelligence and its knowledge on Oracle ERP  PPM , and it was able to read through all the TypeScript code and create descriptions intelligently, which were almost 90% correct when reviewed.

It was very useful because we had 500-plus reusables, and it was able to analyze all of them and put them into a catalog. This makes it very easy to find and use the catalog to determine whether existing functionality is already implemented, preventing redundant implementations.

When it creates a new test, it creates it almost 70 to 80% correctly without errors. The time savings are significant - what previously took one or two days can now be completed in two to three hours maximum. We can complete many more tests in a day or sprint with Cohere's help.

Along with test automation, we handle analysis tasks, and now we have more time for better analysis. We are planning to implement test analysis capabilities as well. Once you receive the requirements and test cases, you can directly use them as input, and it will generate all artifacts and test data.

When performing similarity matching between text descriptions and the catalog descriptions created using Cohere, the matching could be improved. Because it does not have extensive understanding of Oracle functionalities in ERP , it sometimes gives wrong results or the confidence score is lower than desired. Improving that understanding would provide better matches.

When working with Cohere and providing large data sets, there was some hallucination, though it mostly works fine without many issues.

I have been using Cohere for almost seven to eight months.

I have not faced any downtime or related issues. It works fine.

I used Llama but it was not giving results comparable to what I get from Cohere when comparing the two solutions. We only had these two options at that time, and we chose Cohere over Llama.

The setup was pretty smooth. I was able to find things easily. The documents were readily available on the internet, and I was able to find and integrate them without any issues. I subscribed to emails about new model updates, which allowed me to stay current. Oracle has now wrapped it inside their own AI, and we are using the latest version of Cohere as our chosen model.

I started with the public version and then they wrapped it inside Oracle's system. I believe it is private, only accessible to Oracle employees with proper authentication and sign-in details. The pricing and setup were handled by the organization, so I am not aware of those aspects.

We only had two options at that time: Llama and Cohere. After trying both, we chose Cohere over Llama.

Try it and use it. If you find it worthy, then implement it. I have shared all my experiences with you. My rating for Cohere is 7 out of 10.

I use it for a personal project, a Discord bot for my Discord server. I haven't used it that much, but so far it's amazing. I like the support team. They are very good.

Everything is definitely intuitive, and whenever you have an issue, it's very easy to reach out to them on Discord. They're very active, so I'm not really complaining about having issues.

The very first thing that I really like about it is the support team. They're really available on Discord, and they answer all of your questions.

I think it's free for personal projects unless you want to go to production. I haven't really used it that much, but the features that I have used so far, I have no issues with them.

Cohere has text generation. I think it is mainly focused on AI search. If there was a way to combine the searches with images, I think it would be nice to include that.

I've recently started exploring Cohere. It has been a few months now, two to three months.

I'll rate the stability a six out of ten since I haven't been using it much. I haven't really seen any issues.

I will rate the scalability a seven out of ten because I haven't explored it all the way.

The customer service and support are very good.

It's very easy. You just need an API key, and all the configurations are there. It's very easy to start.

If you have worked with something that requires API keys, you should be good to go. I don't think you need a lot of experience.

Cohere has a free tier. You can use the API in development mode, so you can just use it for free. But if you go to production, you will have to pay.

I would advise someone to really consider it first if they really need it because it can be expensive.

So it might be a little expensive.

Overall, I would rate it a seven out of ten.

I would recommend it to others because it is very promising, so it would be worth the time. Others should try it.