Snowflake's data cloud will now include two LLMs trained by Reka: Flash and Core. Flash is a 21-billion parameter model that matches the performance of many larger models across various language and vision benchmarks, while Core is Reka's largest and most performant model. The integration of these models will enable applications such as video captioning, image labeling, e-commerce product description generation, and analytical question answering based on charts. Over 400 enterprises are already using Cortex and the models it hosts to build generative AI applications.
Key takeaways:
- Snowflake Data Cloud is partnering with AI startup Reka to integrate its large language models (LLMs) into its data platform, allowing for generative AI applications that work with text, images, and videos.
- Two LLMs trained by Reka, Flash and Core, will be included in Snowflake's platform. Flash is a 21-billion parameter model, while Core is Reka's largest and most performant model.
- Over 400 enterprises are currently using Snowflake's Cortex and the models it hosts to build generative AI applications in the data cloud, with applications ranging from security vulnerability analysis to healthcare provider chatbots.
- Snowflake's AI innovation pipeline is in 'overdrive', with more AI innovations expected to be announced in the run-up to its annual summit in June.