Mistral AI and its Large model are getting closer to Snowflake


The young French artificial intelligence unicorn has already found two international partners for its new language model called Large – a non-open source solution unlike the previous ones.

In a press release, Snowflake formalizes its partnership with Mistral AI and the availability of Large to Snowflake Data Cloud customers. Access to Large takes place through a global agreement over several years providing at the same time for the publisher’s participation in the unicorn’s A series.

LLM and Data united on Snowflake

The firm thus takes a stake in the startup through its Snowflake Ventures fund, for an unspecified amount. Mistral AI’s Series A fundraising of 385 million euros was announced last December.

“Mistral AI and Snowflake will provide the capabilities businesses need to seamlessly harness the power of large language models (LLMs), while maintaining the security, privacy, and governance of their most valuable asset: their data.” , indicates the American.

Large is not the only model of the tricolor unicorn to now be accessible from the Snowflake environment. The publisher specifies that Mixtral 8x7B and Mistral 7B, open source models, are also available.

This announcement comes as part of the preview launch of Snowflake Cortex, the company’s new managed service dedicated to LLM and vector search – two inseparable building blocks of generative AI.

Vector search and models in managed services

With Cortex, Snowflake promises its users to accelerate their use of analytics and the creation of AI applications exploiting their data stored on its cloud platform – supported by the main cloud providers on the market.

“With Snowflake as their trusted database, we are transforming how enterprises harness the power of LLMs with Snowflake Cortex so they can cost-effectively address new AI use cases within security constraints and confidentiality of the Data Cloud.”

With this statement, Sridhar Ramaswamy, CEO of Snowflake, highlights the synergies between the LLMs provided by Mistral AI and the operational data hosted by its customers on their cloud platform.

Snowflake facilitates its clients’ AI apps

Snowflake is not limited to Mistral models. On Cortex, it includes several LLMs intended for processing specialized tasks, including Llama 2. And to facilitate usage, the publisher has also signed an agreement on inference with Nvidia.

Still with this in mind, the firm is making available (in preview) Snowflake Cortex LLM Functions. The platform thus provides specialized functions allowing the exploitation of LLMs for tasks such as translation, summaries or sentiment analysis.

Mistral AI and Snowflake come together around the ambition described by Arthur Mensch, CEO and co-founder of the French unicorn: “to further democratize AI so that users can create more sophisticated AI applications that generate value at the same time. ‘ladder.”



Source link -97