The cloud computing company Snowflake unveiled the “enterprise-grade” generative AI model, Arctic LLM. It is available under an Apache 2.0 license and is designed for “enterprise workloads,” which include database code generation. It is open source for commercial and research use.

“I think this is going to be the foundation that’s going to let Snowflake and our customers build enterprise-grade products and actually begin to realize the promise and value of AI,” CEO Sridhar Ramaswamy expressed in a press conference. “You should think of this very much as our first, but big, step in the world of generative AI, with lots more to come.”

Similarly, Databricks’ generative AI model, DBRX, is positioned as being specific to the commercial market. Snowflake makes a clear comparison between DBRX and Arctic LLM, stating that DBRX is inferior to Arctic LLM on the two tasks of SQL generation and coding. Arctic LLM outperforms Mistral’s Mixtral-8x7B and Meta’s Llama 2 70B in those tasks.

Additionally, Arctic LLM performs “leadingly” on the widely used MMLU general language understanding benchmark. However, it should be noted that although MMLU aims to assess generative models’ reasoning skills through logic problems, it contains questions that can be answered by rote memory.

Baris Gultekin, head of AI at Snowflake, said “Arctic LLM addresses specific needs within the enterprise sector, diverging from generic AI applications like composing poetry to focus on enterprise-oriented challenges, such as developing SQL co-pilots and high-quality chatbots.”

According to Snowflake officials, it was able to train Arctic LLM at “roughly one-eighth the cost of similar models” using public web data sets such as RefinedWeb, C4, RedPajama, and StarCoder because of its effective design.