Snowflake, the data cloud company, is partnering with AI startup Reka to enhance its data platform with multimodal large language models (LLMs). This collaboration will enable enterprises to build generative AI applications that can work with text inputs, images, and videos. Snowflake had previously participated in Reka’s funding round last year. The addition of Reka’s LLMs, including the Flash and Core models, will allow Snowflake users to develop applications for various use cases such as captioning videos, labeling images, generating product descriptions, and answering analytical questions based on charts.
Snowflake’s journey towards becoming the data infrastructure of choice for its customers has not been easy. It initially provided users with a simple data warehouse but later expanded its capabilities to support more data formats and build a full-fledged data cloud. Last year, it launched Snowflake Cortex, a fully managed service designed for building LLM apps, to cater to the growing demand for generative AI. Cortex offers a suite of AI building blocks, including open-source LLMs, to analyze data on the platform and develop applications for different business-specific use cases.
Reka Flash, a 21-billion parameter model, is optimized to deliver high performance across various language and vision benchmarks. On the other hand, the Core model is Reka’s largest and most performant model, comparable to GPT-4 and Gemini Ultra in terms of performance. Snowflake plans to add the Flash model to Cortex immediately and is developing support for the Core model to ensure its availability soon. The company may also include other models from Reka based on demand.
The integration of Reka’s AI models into Cortex will enable Snowflake users to create generative AI applications that go beyond text inputs. This will be beneficial for various use cases such as chatbots that understand charts and entertainment businesses generating marketing content from video or image assets. Over 400 enterprises are already leveraging Cortex and the models it hosts to build generative AI applications in the data cloud. The addition of Reka’s models will bring the total number of LLMs on the platform to twelve, including models from Mistral and Google.
Snowflake’s vision is to make AI accessible to all users and deliver recognizable business value faster. The company aims to enhance productivity, collaboration, and accelerate AI and machine learning workflows for its customers. Snowflake’s roadmap is built on the principle that there is no AI strategy without a data strategy, emphasizing the importance of data as the fuel for AI. The company is constantly innovating and will be announcing more AI advancements leading up to its annual summit in June.
Snowflake’s competitor, Databricks, has also been introducing open models and strengthening its tooling to build generative AI applications. Last week, Databricks acquired Lilac, an AI startup that helps teams analyze and fix unstructured data for AI training, further expanding its capabilities in the AI space.
In conclusion, Snowflake’s collaboration with Reka to enhance its data cloud with multimodal LLMs will empower enterprises to build generative AI applications that can process text inputs, images, and videos. This partnership strengthens Snowflake’s position as a leading provider of data infrastructure and demonstrates its commitment to delivering AI innovations to its customers. With the addition of Reka’s models, Snowflake users will have access to a wider range of AI capabilities, enabling them to drive value from their datasets in new and innovative ways.