Transform structured data with SAP-RPT-1.5

Operationalize SAP’s relational pretrained transformer model to gain predictive, explainable insights from enterprise data without building models from scratch.
Man on a laptop working, with images of the user interface of the SAP RPT-1 playground and testing environment for predictive insights with tabular AI on structured data.

What is SAP-RPT?

SAP-RPT is a fully managed relational pretrained transformer model that uses in-context learning to deliver accurate, explainable predictive insights from structured business data.

placeholder

The SAP-RPT-1.5 AI model, which is now generally available, includes:

  • Pretraining for structured data to eliminate costly and time-consuming model training.

  • In-context learning that delivers real-time, dependable predictions amid changing and incomplete data.

  • Built-in governance, reliability, and explainability that enable end-to-end pipeline management with.

  • Two commercial versions in the generative AI hub, each tuned for a different speed and accuracy balance.

Key benefits

Accelerate time to insights

Deploy predictions instantly, reducing weeks of data preparation to hours, without model training or specialized ML expertise.

Generate accurate insights from imperfect data

Achieve up to 2X prediction quality compared to narrow AI models and 3.5X compared to language models, even when business data is incomplete or changing. SAP-RPT’s resilient architecture, in-context learning, and explainable insights help identify the signals driving outcomes while addressing data-quality challenges.*

Scale predictive power across the enterprise

Use a single model for classification and regression across business units and scenarios, simplifying deployment and reducing the need for multiple use-case-specific predictive models.

Key features

placeholder

In-context learning API

SAP-RPT-1.5 processes classification and regression tasks using existing business data sources or sample records provided directly in API calls. The model identifies data patterns within each request’s data context to generate accurate predictions.

Pretrained relational architecture

Built with an inherent understanding of tabular structures, data relationships, and business logic, SAP-RPT-1.5 handles structured datasets, changing values, and enterprise data formats without custom preprocessing or feature engineering.

Tabular orchestration workflows

Operationalize SAP-RPT-1.5 with a unified enterprise-grade platform that embeds predictive intelligence into pipelines and manages them with built-in governance, reliability, and explainability.

Two versions for infinite use case possibilities

SAP-RPT-1.5

  • Low latency.

  • High prediction throughput.

  • Fully capable of handling most predictive scenarios.

SAP-RPT-1.5-large

  • High accuracy.

  • Lowest error rates.

  • Complex prediction use cases.

Resources

placeholder

ConTextTab: Semantics-aware learning for tabular data

ConTextTab unites language model semantics with table-native efficiency to achieve high performance on real business tabular prediction tasks.

placeholder

In-Context learning for relational data

Read more about SAP's revolutionary table-native foundation model that uses in-context learning to handle enterprise prediction tasks without custom training.

FAQ

An RPT model is an AI system pre-trained on large, structured, relational datasets to learn common tabular data patterns. Unlike large language models that process text sequences, SAP-RPT-1.5 is designed to understand relationships in structured data, such as sales records, financial transactions, supplier performance tables, or human resources metrics.

In-context learning allows the model to understand patterns on the fly and instantly adapt to new scenarios from example records in your API call to provide immediate and accurate contextual predictions.

Yes. SAP-RPT-1.5 can be tested for free at the SAP-RPT playground at https://rpt.cloud.sap. Configure prediction scenarios, evaluate model performance, generate predictions, and change and interact with data in natural language. This model provides insights into data changes and surfaces relationships with your data. Explore model-detected statistical patterns through an intuitive interface for testing multi-class classification and numerical regression using your CSV files or SAP-provided samples.

Yes. An open-source version, sap-rpt-1-oss, is available on Hugging Face and GitHub for research and non-commercial purposes. SAP-RPT-1.5 is only available in SAP's generative AI hub in two commercial versions.

The tabular orchestration workflow capability (generally available in late 2026) will operationalize the model. The capability will be delivered on SAP’s generative AI hub, accessible via APIs, and launched with context-selection and model-selection modules. The workflow will remove the need for manual context engineering and model handling by automatically selecting the right context, building the API call, and returning predictions based on query rows provided at runtime.

Yes, model observability (generally available in July 2026) will enhance trust and transparency through predictions, detailed factor impact, and data attribution—complementing the top-k predictions and confidence interval enhancements already in SAP-RPT-1.

Yes. The model previously referred to as the -small version is now simply SAP-RPT-1.5, our standard SAP-RPT model, which is fully capable of handling most predictive scenarios, offering low latency, high speed, and strong accuracy.

*Numbers assumed for consumer products company with €1 billion in revenue and 2,000 employees. Annual benefits result from one or more value drivers. Source: SAP Value Management.
**Legal disclaimer: Token and after-promotion optimization costs apply

twitter pixeltwitter pixeltwitter pixeltwitter pixeltwitter pixeltwitter pixeltwitter pixeltwitter pixeltwitter pixeltwitter pixeltwitter pixeltwitter pixel