SAP-RPT (RPT-1 → 1.5 → 1.6) — Using SAP's Tabular Foundation Model

SAP-RPT (pronounced "rapid") is SAP's family of tabular foundation models: pretrained models that predict a value in a business table — a class such as a cost centre or a payment-delay bucket, or a number such as days late or spend — from the rows you send with the request, without training or fine-tuning a model for the use case. This page covers the product side: versions, how you get access, how you call it, the limits, the cost mechanics and where it fits in a delivery. How the model works internally is covered in C237.

Versions and status (as of 23 September 2026)

• SAP-RPT-1 — announced at SAP TechEd in November 2025 and generally available in the generative AI hub of SAP AI Core since 8 December 2025, as sap-rpt-1-small (speed, throughput) and sap-rpt-1-large (accuracy). A later update added a top_k input and confidence intervals in regression output. SAP has announced that SAP-RPT-1 will be retired by 31 December 2026.
• SAP-RPT-1.5 — generally available on 27 July 2026, as sap-rpt-1.5 and sap-rpt-1.5-large, with an explainability feature.
• SAP-RPT-1.6 — generally available on 7 September 2026, as sap-rpt-1.6 and sap-rpt-1.6-large, with explainability and a new context_mode parameter on the large variant.
• SAP-RPT-1-OSS — an open-source variant on Hugging Face and GitHub, for research and non-commercial use; it is not the commercial model.

SAP's product page now calls the former "small" model simply the standard SAP-RPT model. All What's New entries flag the models as not available in sovereign cloud offerings.

Why it matters

• The earlier 'research artefact, not a product' framing is obsolete: RPT has been GA in the generative AI hub since 8 December 2025 and is on its third release (1.6, 7 September 2026).
• It turns a tabular prediction from a model-building project into an API call, which changes how you scope and price predictive use cases in SAP programmes.
• RPT-1 retires by 31 December 2026 — any pilot or proposal still on sap-rpt-1-* needs a migration line item now.

Key points

• Versions: RPT-1 (GA 8 Dec 2025, retiring by 31 Dec 2026), RPT-1.5 (GA 27 Jul 2026), RPT-1.6 (GA 7 Sep 2026); each in a standard and a -large variant; RPT-1-OSS is research-only.
• Access: free RPT Playground (rpt.cloud.sap, 25 rows / 4 columns / 50 classes; API up to 1,000 requests/hour) or the generative AI hub in SAP AI Core on the extended plan.
• Call pattern: context rows with known targets, then query rows with [PREDICT]; POST to /v2/inference/deployments/<id>/predict with prediction_config, index_column and rows.
• RPT-1.6 limits: 2,048 vs 65,536 context rows, 100 vs 256 columns, 512 query rows and 10 target columns per request; deep context mode only on -large above 8,000 context rows.
• Tabular orchestration registers data destination, tabular artifact and scenario once, then predicts from query rows only; cost is metered in GenAI tokens → capacity units (SAP Note 3437766).
• Since SAP's July 2026 Prior Labs acquisition, RPT shares the generative AI hub with Prior Labs' TabPFN-3.5-Plus (GA 15 Sep 2026); tabular orchestration is planned to route to both, so the choice becomes model-per-table-shape, benchmarked, not vendor loyalty.
• Decide standard vs -large vs deep mode by table complexity and available context volume; decide RPT vs an incumbent GBDT by how often the use case changes and whether a data-science team already owns it.

Terms used on this page

Context rows

Rows sent in the request whose target value is known; the model learns the pattern from them at inference time.

Query rows

Rows whose target cell holds the [PREDICT] placeholder; they must follow the context rows.

Generative AI hub

Capability of SAP AI Core and SAP AI Launchpad that exposes foundation models, including SAP-RPT, under the extended service plan.

Deep context mode

context_mode "deep" on sap-rpt-1.6-large: more context rows used for higher accuracy, at higher latency and cost.

Tabular orchestration

Managed generative AI hub workflow that stores data destination, tabular artifact and scenario once and selects context rows at each /v1/predict call.

Sources

1. SAP News — Business AI innovation unveiled at SAP TechEd (Nov 2025: SAP-RPT-1, playground, OSS)
2. SAP Help — SAP AI Core guide incl. What's New (RPT-1 GA 2025-12-08, RPT-1.5 2026-07-27, RPT-1.6 2026-09-07; metering)
3. SAP Help — SAP-RPT-1.6 model specifications and context mode
4. SAP Help — SAP-RPT-1.5 model specifications
5. SAP Developers — Strategic Prediction with SAP RPT-1 (tutorial: deployment IDs, payload, SDKs)
6. SAP Community — SAP-RPT-1.6, Tabular Orchestration and RPT Playground API now available (Sep 2026)
7. SAP — SAP-RPT product page (versions, FAQ, model cards)
8. SAP Community — From Data to Prediction: Integrating RPT-1.5 into SAP Analytics Cloud (Aug 2026)
9. SAP Community — Calling SAP HANA Cloud ML and RPT-1.5 from SAP Snowflake via External Access Integration (Aug 2026)
10. SAP Architecture Center — Predictive & Tabular AI (golden path)
11. SAP Community — SAP-RPT-1.5 is Now Generally Available (July 2026)
12. SAP News Center — TabPFN-3.5-Plus Now Available in SAP AI Core (15 September 2026)