Snowflake beats Databricks to integrating Claude 3.5 directly

Soon after announcing its earnings for the third quarter of FY25, Snowflake dropped a bombshell: It is teaming up with leading AI vendor Anthropic to further advance AI projects for its customers. 

The data ecosystem giant has signed a multi-year strategic agreement, bringing Anthropic’s Claude 3.5 family of models to its core platform on AWS. The move will allow enterprises using Snowflake to leverage the cutting-edge Claude large language models (LLMs) for building advanced AI applications, including conversational agents. The engagement will also power Snowflake’s customer-facing agentic offerings as well as help the company’s employees accelerate their internal workflows.

The partnership comes as the latest in a series of efforts by Snowflake to provide enterprises with everything they need to develop secure and governed generative AI applications using their data hosted on the company’s platform. Just recently, it even announced observability capabilities, enabling users to keep a close eye on the performance of the LLM applications they develop on their Snowflake instance. 

Interestingly, this strategy is similar to that of Databricks, Snowflake’s biggest competitor, which also offers a range of open-source and proprietary models. In fact, the Ali Ghodsi-led company already allows customers (in its own way) to integrate proprietary external models, including those from Anthropic.

Claude 3.5 family directly available on Snowflake

Last year, Neeva co-founder Sridhar Ramaswamy took over as the CEO of Snowflake, marking a new chapter of AI resurgence for the data company. It began dropping new AI capabilities, allowing customers to not only accelerate their data workflows on the platform (like with text-to-SQL) but also enable powerful AI-based use cases. One such offering that made headlines last year was Cortex, a fully managed service to build LLM apps.

Cortex provides enterprises using data cloud with a suite of AI building blocks, including multiple LLMs, to analyze data they have on the platform – with the same security and privacy of Snowflake – and build applications targeting different business-specific use cases. The company started with LLMs for specialized tasks such as sentiment analysis and soon had a roster of several open models, including those from Google, Meta and Mistral, and a few proprietary ones.

Now, as the next step in this work, the company has partnered with Anthropic to bring its proprietary Claude 3.5 family of models to Cortex AI. The move is significant as it allows Snowflake’s 10,000+ customers to power their gen AI app experiences with Claude’s reasoning, planning and problem-solving. 

For instance, Claude 3.5 Sonnet, the most intelligent model in the family, already outperforms OpenAI’s GPT-4o and Google’s Gemini 1.5 Pro in key areas such as graduate-level reasoning, code generation, multilingual math and grade school math. These capabilities can come in very handy for agents aimed at analyzing data, running ad-hoc analytics, generating visualizations and executing other multi-step workflows. 

“By bringing Anthropic’s industry-leading models to customers’ enterprise data where it already lives, within the security and governance boundaries of the AI Data Cloud, we will unleash new ways for businesses to harness this data for agentic use cases, coding assistants, document chatbots, unstructured data analytics, and more,” Christian Kleinerman, EVP of product at Snowflake, said in a statement.

It’s worth noting here that while Databricks already allows users to leverage the entire range of Claude models for gen AI application development, the company does not host the models directly in its platform. They are only available via a REST API. This partnership, on the other hand, ensures that Snowflake customers can access Claude models (just the latest 3.5 series) directly within their data cloud instance, right where their structured and unstructured datasets reside. 

“Snowflake’s approach to AI is distinct in that we offer enterprises a unified, simple platform experience where everything just works and is available with Snowflake’s security and governance perimeter. Compared to other vendors that require painstaking configurations, Snowflake customers save valuable time and money on administrative tasks so that they can spend it on what matters most: launching projects and products faster for end users,” Harshal Pimpalkhute, principal product manager at Snowflake, told VentureBeat.

Claude to power internal workflows too

Beyond expanding the ecosystem of LLMs on Cortex AI, this partnership will also see Snowflake leveraging the Claude 3.5 family of models for its internal workflows and customer-facing products.

For customer products, the company says it will use Claude as one of the key models powering its new agentic offerings, including recently-announced Snowflake Intelligence as well as Cortex Analyst. The products will be optimized for the new LLMs, which will drive faster access to downstream results with high levels of accuracy.

“Claude’s industry-leading accuracy and expansive context window make it an ideal fit to power Snowflake’s agentic products such as Snowflake Intelligence, which enables enterprises to easily ask business questions across their enterprise data to unlock data-driven answers, and in just a few steps, create data agents that take action on those insights,” Pimpalkhute said.

As for internal use cases, the product manager noted that Snowflake employees will be using Claude models to create custom agentic workflows. The deployment hasn’t happened yet, but the company soon hopes to leverage it as a way to increase employee efficiency, allowing them to focus on more high-value tasks. 

Currently, more than a thousand enterprises – over 10% of Snowflake’s total customer base – use Cortex AI for developing gen AI applications. The addition of Anthropic’s models to the service is expected to improve the value proposition of the company’s data cloud, drawing more customers to it.