Early Preview

Gretel's Tabular LLM

Gretel's novel tabular large language model is capable of generating highly valuable synthetic tabular data. Trained on half a trillion tokens, our model has captured the knowledge of most of the datasets on the public internet.
Request early access

Why try Gretel’s Tabular LLM?

  • Generate data from scratch

    Quickly generate highly realistic data for demos, testing, and pre-production environments. Create domain-specific data for any region, in any language.

  • Generate data for existing datasets

    Have incomplete data? Edit and augment existing datasets, and fill in gaps where data is missing. Use both SQL and natural language for prompting.

  • Zero-shot prompting. Generate correlated datasets without needing any training data.

  • Augment. Automatically add columns and records to existing data.

  • Fill-in gaps. Replace nulls and missing fields with realistic values.

  • SQL. Use a SQL schema or query to create an entire dataset with the correct datatypes.

  • Regionalize. Build datasets for different regions, including names, email domains, addresses, and company names.

  • Operationalize. Generate millions of records and integrate them into your MLOps pipelines.