Dataster Documentation

LLMs Overview

When building a GenAI application, developers often need to test multiple models to determine which one performs best for their specific use case. To facilitate this, Dataster supports over 15 curated, high-quality Large Language Models (LLMs) from various providers across several cloud platforms. The full list of models can be found on our pricing page here.

Models in Dataster fall into two main categories: completion and embedding. For convenience, Dataster provides out-of-the-box access to each of the supported models. These models are automatically populated in the LLM Catalog upon account creation. A filter allows you to view only the LLMs from a specific cloud provider or those with a specific pattern in the name, such as "oai".

The models that Dataster provides access to out-of-the-box are shared. For private access, users can bring their own models into Dataster as assets, ensuring that only they can access those models. The catalog displays Bring Your Own (BYO) LLMs differently from the Dataster-provided LLMs. Instead of showing the Dataster badge, it gives the user the option to edit or delete the model from the catalog.

Conclusion

Dataster supports over 15 Large Language Models for completions or embeddings. Shared access to each model is provided out-of-the-box to every user. Users are also welcome to bring their own models into the LLM Catalog for private access, isolated from other Dataster users. Models added to the catalog can later be used for use case-specific evaluation.

If you encounter any issues or need further assistance, please contact our support team at support@dataster.com.