Perhaps no executive in the artificial intelligence field has gone to greater lengths to knock down generative AI hype quite like Databricks CEO Ali Ghodsi. A year ago, that looked like sour grapes as OpenAI’s quick progress made many people think that software tools like those Databricks sells to companies to manage data and utilize AI would soon be obsolete.

It hasn’t turned out that way, and Ghodsi has been proven mostly right so far. ChatGPT is a big business, but generative AI isn’t a magic wand. Using it to power automated customer service or data analysis is a grind, requiring precision. Some cloud providers are tempering expectations around how much corporations will spend on the technology in the near term. (See recent stories here and here.) Meanwhile, Databricks’ revenue is growing just fine. Ghodsi, it should be noted, has also made some bold proclamations about the price of Nvidia’s AI chips coming down.

So we were surprised this week to see Databricks, now valued at $43 billion, join the race to develop the best open-source large language model by releasing one called DBRX. (You can check out the blog post on the release here.) Meta Platforms and Mistral have dominated the open-source LLM field, providing an alternative to OpenAI’s closed-source models that developers pay to use. (Mistral actually used Meta’s model to build its own, whereas Databricks did not rely on the work of competitors, according to Jonathan Frankle, Databricks’ chief scientist.)