Databricks invests in Mistral, announces AI tie-up

by DTM | Posted on

Databricks has announced a new strategic partnership and investment in Mistal AI, one of the most intriguing generative AI start-ups around today.

The pair say their collaboration will look to integrate Mistal AI’s open models into the Databricks Data Intelligence Platform, offering customers enhanced accessibility and flexibility.

The news comes shortly after Microsoft Vice Chair and President Brad Smith confirmed a “multiyear partnership” between the Redmond company and Mistral AI.

Databricks and Mistral AI

Mistral’s models are now natively integrated within the Databricks platform, which means that customers will be able to browse and use them in the Databricks Marketplace.

“We are delighted to forge this strategic alliance with Databricks, reaffirming our shared commitment to the portability, openness and accessibility of generative artificial intelligence for all,” noted Mistral AI CEO Arthur Mensch.

Two flagship models were introduced as part of the collaboration: Mistral 7B and Mistral 8x7B. The former is a dense transformer model trained with 8k context length and seven billion parameters.

Mistral 8x7B is a sparse mixture of experts model (SMoE) trained with a context length of 32k and 45 billion parameters. It’s also capable of handling English, French, Italian, German, and Spanish.

The French startup’s models can be found on Databricks’ Marketplace, where they can be deployed with Mosaic AI Model Serving, batch inferenced with Spark, and model inferenced in SQL.

Prem Prakash, Head of AI/ML Product Marketing at Databricks, added: “Now, we’re offering a streamlined experience that will make it even easier for customers to natively access, customise, deploy and govern Mistral AI models for their generative AI applications.”

Precise details of the investment have been kept under wraps, but the fact that Mistral AI has already attracted attention from Microsoft, a company that has poured billions into OpenAI, is significant news.

More from TechRadar Pro