Large Tabular Models: Fundamental raises $255M to build models for enterprises (techcrunch.com)

Fundamental, an AI lab newly emerged from stealth, has announced the launch of Nexus, a large tabular model (LTM) designed to extract insights from the vast amounts of structured data generated by enterprises. With $255 million in funding at a $1.2 billion valuation, led by prominent investors including Oak HC/FT and Salesforce Ventures, Fundamental aims to redefine how large organizations analyze data. CEO Jeremy Fraenkel highlighted that while large language models (LLMs) excel with unstructured data, they struggle with structured data like tables, which is essential for effective enterprise operations. Nexus distinguishes itself by being deterministic, providing consistent responses to inquiries, and it deviates from the prevalent transformer architectures seen in contemporary AI. This model is built to navigate the complexities of large datasets—such as spreadsheets with billions of rows—where traditional transformer-based models often face limitations due to their context window size. By applying advanced techniques to big data analysis, Fundamental promises to enhance performance across various use cases, potentially transforming how businesses manage data resources. The company has already secured significant contracts, including agreements with Fortune 100 clients and a strategic partnership with AWS for seamless deployment of Nexus.