Snowflake Adds AI21 Labs’ Jamba-Instruct LLM to Cortex AI Service for Enterprise Customers

Snowflake, the data cloud giant, has announced its partnership with Israeli AI startup AI21 Labs to integrate the Jamba-Instruct LLM into its Cortex AI service. This move will allow Snowflake’s enterprise customers to build generative AI applications capable of handling long documents without compromising quality and accuracy. Jamba-Instruct is not Snowflake’s only large language model (LLM) partner. The company has been actively engaging with various partners to create a whole ecosystem for developing high-performing, data-driven AI applications.

Jamba-Instruct, an instruction-tuned version of AI21’s Jamba model, offers several benefits to Snowflake users. It provides access to a large context window of up to 256K tokens, equivalent to approximately 800 pages of text. This makes it a powerful model for extensive document processing, such as analyzing corporate financial history or clinical trial interviews. Financial analysts, clinicians, and retailers can benefit from the model’s capabilities to extract relevant information, build summarization tools, and create chatbots capable of sustaining coherent and reference-based conversations with customers.

One of the major advantages of using Jamba-Instruct is its cost-effectiveness. The hybrid nature of the model, combined with mixture-of-experts (MoE) layers, makes its 256K context window more economically accessible than other instruction-tuned transformer models of the same size. Additionally, Cortex AI’s serverless inference with a consumption-based pricing model ensures that enterprises only pay for the resources they use, eliminating the need for dedicated infrastructure.

Snowflake’s Cortex AI service already covers a range of LLMs, including its own Arctic model, as well as models from Google, Meta, Mistral AI, and Reka AI. The company aims to provide customers with flexibility in choosing the best model for their specific use case, cost, and performance requirements. Snowflake listens to customer feedback and continuously evaluates and integrates LLMs that address specific needs and use cases.

To assist customers in navigating the growing number of model choices, Snowflake acquired TruEra, a startup specializing in model evaluation. Customers can use TruEra’s TruLens offering to run LLM experiments and determine the most effective models for their needs.

Currently, over 5,000 enterprises are utilizing Snowflake’s AI capabilities, with the top use cases being automated business intelligence, conversational assistants, and text summarization. As Snowflake expands its model offerings, it aims to cater to a broad range of use cases and ensure that each model has unique capabilities. Jamba-Instruct, with its large context window, offers a distinctive advantage in the market.