Data cloud provider Snowflake has launched an open source large language model, Arctic LLM, as part of a growing portfolio of AI offerings that help businesses leverage their data. Typical use cases include data analytics, including review sentiment analysis, chatbots for customer service or sales, and business intelligence queries, such as extracting revenue information.
Snowflake's Arctic is offered alongside other LLM models from Meta, Mistral AI, Google and Reka in its Cortex product, which is only available in select regions. Snowflake said Cortex will be available in APAC in Japan in June through the AWS Asia Pacific (Tokyo) region. The offering is expected to roll out to customers globally and the rest of APAC over time.
Arctic will also be available through the Amazon Web Services hyperscaler, as well as other model gardens and catalogs used by companies, including Hugging Face, Lamini, Microsoft Azure, NVIDIA API Catalog, Perplexity, Together AI and others, according to the company.
What is Arctic Snowflake?
Arctic is Snowflake's new “next-generation” LLM, launched in April 2024, designed primarily for enterprise use cases. The company has shared data showing that Arctic scores well against other LLMs on several benchmarks, including SQL code generation and instruction following.
Snowflake's head of AI, Baris Gultekin, said the LLM took three months to build (one-eighth the time of other models) on a $2 million budget. This achievement takes the model beyond how quickly and cheaply an enterprise-level LLM can be developed.
What are Snowflake Arctic's key differentiators?
The goal of Arctic LLM is to provide “efficient intelligence”; excels at common business tasks while being more economical to use when training custom AI models on business data. It is also pushing the boundaries of open source as it was released with an open source Apache 2.0 license.
Rather than taking the general-purpose world understanding offered by many other open source LLMs, including the Llama de Meta models, the Arctic AI model aims to specifically meet enterprise demand for “conversational SQL data co-pilots, RAG code and chatbots”.
SEE: Zetaris on federated data lakes and the enterprise data mess
“Business intelligence” capabilities
Snowflake created its own “business intelligence” metric to measure LLM performance, which was a combination of coding, SQL generation, and instruction following capabilities.
Arctic fared favorably against models from Databricks, Meta, and Mistral in common AI model benchmarking tests, which challenge and provide a percentage score for LLM models in specific capability domains. According to Snowflake, the model's ability to excel in business intelligence when compared to LLMs with higher budgets was notable.
Training and inference efficiency.
Gultekin said Arctic AI LLM offers enterprise clients a way to train custom LLMs using their own data in a more cost-effective way. The model is also designed to perform efficient inference to make enterprise implementations more practical and lower cost.
Open source with Apache 2.0
Snowflake making Arctic LLM open source with an Apache 2.0 license is partly due to what Gultekin says is the AI team's deep experience in open source. This means that the company provides access to weights and codes, as well as data recipes and research insights.
Snowflake believes the industry and the product itself will be able to move faster through genuine contributions from open source developers, while Gultekin said being able to see under the hood would help enterprise customers trust the model more.
How will Snowflake Arctic affect the AI market?
Snowflake's launch in Arctic made waves in the enterprise data and technology community, thanks to its speed, efficiency, and SQL generation capabilities. Gultekin said the company's decision to “go beyond open source” has generated excitement in the research community.
SEE: Our comparison of Snowflake with Azure Synapse Analytics
“This is our first release and it sets a really good benchmark. The market will evolve in such a way that there will not be a single winner; Instead, all customers are very interested in choosing in the market. We have already seen great usage and we hope that continues,” he said.
Does Snowflake have AI experience?
Snowflake previously offered a number of machine learning solutions. As part of the generative AI boom in 2023, it acquired several AI organizations, including data search company Neeva and NXYZ, a company of which Gultekin was CEO and co-founder. Snowflake has since developed its core generative AI platform, AI search capabilities, and is now adding LLM models.
