Large Tabular Models NEXUS from Fundamental: LLMs Surrender to Tables

On February 5, 2026, startup Fundamental emerged from stealth mode with $275 million in funding. The company unveiled NEXUS — the first Large Tabular Model (LTM), specifically built to analyze structured data in tables.

Why LLMs Don't Work with Tables

Models like ChatGPT, Claude, and Gemini produce convincing text and images, but ask them to analyze even a small table — they look lost.

• LLMs are designed for sequential data: each word influences the next, order is critical
• Tables don't require order: you can rearrange columns or rows — the meaning doesn't change
• LLMs output different answers with minimal input changes (usually useful), but for fraud detection you need determinism
• For most companies, the most important data lives in spreadsheets — bank transactions, sales logs, website metrics

How NEXUS Works

Unlike LLMs, which model sequences of tokens, LTMs directly model table structure. The model learns from three sources simultaneously: numerical value, its meaning, and statistical relationships with other data in the column.

Companies need deterministic, consistent predictions.

With LLMs, this is impossible, — explains Jeremy Fraenkel, CEO of Fundamental.

Fundamental trained NEXUS on billions of tables: a combination of licensed datasets, open sources, and synthetic data. The company owns proprietary datasets through partnerships and does not train on client data. NEXUS runs on confidential computers — Fundamental physically cannot access user data.

Who Else Is Developing LTMs

In June, Amazon embedded NEXUS into SageMaker — the standard machine learning tool on AWS. This gives the model access to sensitive client data that companies manage locally.

In March 2026, companies Feedzai (financial security) and Mastercard launched their own LTMs to fight fraud. At the end of June, Google introduced TabFM, trained on hundreds of millions of synthetic tables. Scientists developed FlexTab, TabICL, and iLTM — LTMs are becoming a trend.

What It Means

Tables are the language of business. For 50 years companies relied on XGBoost and other gradient-boosted decision trees for analysis, which had to be manually tuned for months for each task. LTMs promise to make tabular analysis as automated as text generation. LLMs and LTMs together can create more powerful AI — left brain for text, right brain for numbers.