SAP RPT-1 is a relational foundation model for predicting outcomes from tables, not a chatbot or a replacement for every machine-learning system. It can perform classification and regression using in-context examples, potentially reducing the need to train a separate model for each business task. The important qualification is that “no fine-tuning” removes one part of the workflow—not data preparation, validation, governance, monitoring, or operating cost.
What SAP RPT-1 is
RPT stands for Relational Pre-trained Transformer. SAP describes RPT-1 as a table-native relational foundation model built for structured and relational business data. Typical inputs include customer, supplier, transaction, product, payment, or operational records. Typical outputs are a category or a numeric prediction.
Examples include predicting whether an invoice will be paid late, scoring supplier risk, estimating churn, classifying a transaction, or predicting a numeric business outcome. SAP’s goal is a reusable predictive engine that can support many tasks instead of requiring a new narrow model for every use case. See SAP’s announcement at SAP TechEd.
RPT-1 is not an LLM
| System | Input | Typical output | Best fit |
|---|---|---|---|
| Large language model | Text, code, documents | Text or structured response | Dialogue, drafting, extraction and language reasoning |
| Conventional narrow ML | Task-specific structured data | Prediction or classification | One defined business problem |
| RPT-1 | Tables and relational business data | Classification or regression | Multiple structured-data prediction tasks |
An LLM predicts the next token in a text sequence. RPT-1 instead uses known rows and columns to infer a missing or target field. Calling it an LLM obscures both its strengths and its limits: it is not a general-purpose assistant for summarizing documents, generating code, analyzing images, or holding conversations.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitches#1 Best Overall
- 【Leading AI Mini Workstation】MINISFORUM AI MS-S1 Max Workstation comes with AMD Ryzen AI Max+ 395 processor, which uses AMD's latest generation Zen 5 architecture. It has 16 Cores and 32 Threads, the boost clock is up to 5.1GHz. The overall processor performance is up to 126 TOPS, and the NPU performance reaches up to 50 TOPS. AMD Ryzen AI enables improved productivity, advanced collaboration, and improved efficiency.
- 【AMD Radeon 8060S Graphics 】The MS-S1 Max Mini PC equipped with AMD Radeon 8060S Graphics which built on the new generation of RDNA 3.5 architecture AMD graphics, it brings ultra-high frame rate experiences and advanced content creation features anywhere and delivers staggering performance. It can handle all your computing and multimedia tasks efficiently.
- 【Five 8K Video Output】This MS-S1 Max Workstation comes with five video outputs, 1x HDMI (8K@60Hz), 2x USB4(40Gbps,Alt DP2.0,PD out 15W) and 2x USB4 V2(80Gbps,Alt DP2.0,PD out 15W) Outputs, which support multiple monitors display at the same time and provide a larger and wider filed of view and improve your work efficiency. It is used in fields that require high-performance computing and graphics processing, including digital signage and securities trading, as well as work that uses CAD, such as engineering design, scientific calculations, animation production, and post-production for movies and television.
- 【 Fast and Stable Wire & Wireless Speed】It comes with Two 10G Lan Ports for wired connection and and Wi-Fi 7 / BT5.4 for wireless connection, which increased the network speed greatly and expand its functions and improved performance of computer to a large extent and allows you to use more networks such as software routers (OpenWRT / DD-WRT / Tomato etc.), firewalls, NAT, network isolation etc.
- 【Large Storage & Flexible Expandability】This Workstation equipped with 128GB LPDDR5-8000MHz + 2TB M.2 2280 PCIe4.0 SSD. There is another PCIe4.0 SSD slot available for up to 8TB, these SSD slots are compatible with RAID0 and RAID1, you can store movies, videos, photos, important files easily. What’s more, it also comes with 1x standard PCIex16 slot(PCIe4.0x4) inside.
What “no fine-tuning” really means
The model is pre-trained before a customer uses it. For a prediction, the customer supplies a table with input columns and representative examples containing target values; target values in new rows are then marked for prediction. This is in-context learning rather than training a new model for every target.
A simplified pattern might look like this:
PRODUCT PRICE ORDERDATE ID CATEGORY
Couch 999.99 28-11-2025 35 [PREDICT]
Office Chair 150.80 02-11-2025 44 Office Furniture
Server Rack 2200.00 01-11-2025 104 Data InfrastructureThe table still has to be designed correctly. Teams must define the target, select representative examples, resolve missing and duplicate records, prevent leakage, test on unseen data, and monitor drift. SAP explicitly advises testing with production data because the best context length, runtime, cost, and quality vary by dataset (SAP documentation).
What RPT-1 can and cannot predict
Supported core tasks
- Classification: assign a class such as high-risk, late-payment, or churn-likely.
- Regression: predict a numeric value such as an amount, score, or business measure.
A sales forecast might be implemented as regression, but that does not make RPT-1 a universal time-series forecasting system. The cited SAP material does not establish it as a natural-language reasoning, document-understanding, image, speech, ranking, survival-analysis, optimization, or causal-inference model.
Rank #2
- 【Leading AI Mini Workstation】MINISFORUM AI MS-S1 Max Workstation comes with AMD Ryzen AI Max+ 395 processor, which uses AMD's latest generation Zen 5 architecture. It has 16 Cores and 32 Threads, the boost clock is up to 5.1GHz. The overall processor performance is up to 126 TOPS, and the NPU performance reaches up to 50 TOPS. AMD Ryzen AI enables improved productivity, advanced collaboration, and improved efficiency.
- 【AMD Radeon 8060S Graphics 】The MS-S1 Max Mini PC equipped with AMD Radeon 8060S Graphics which built on the new generation of RDNA 3.5 architecture AMD graphics, it brings ultra-high frame rate experiences and advanced content creation features anywhere and delivers staggering performance. It can handle all your computing and multimedia tasks efficiently.
- 【Five 8K Video Output】This MS-S1 Max Workstation comes with five video outputs, 1x HDMI (8K@60Hz), 2x USB4(40Gbps,Alt DP2.0,PD out 15W) and 2x USB4 V2(80Gbps,Alt DP2.0,PD out 15W) Outputs, which support multiple monitors display at the same time and provide a larger and wider filed of view and improve your work efficiency. It is used in fields that require high-performance computing and graphics processing, including digital signage and securities trading, as well as work that uses CAD, such as engineering design, scientific calculations, animation production, and post-production for movies and television
- 【 Fast and Stable Wire & Wireless Speed】It comes with Two 10G Lan Ports for wired connection and and Wi-Fi 7 / BT5.4 for wireless connection, which increased the network speed greatly and expand its functions and improved performance of computer to a large extent and allows you to use more networks such as software routers (OpenWRT / DD-WRT / Tomato etc.), firewalls, NAT, network isolation etc.
- 【Large Storage & Flexible Expandability】This Workstation equipped with 64GB LPDDR5-8000MHz + 2TB M.2 2280 PCIe4.0 SSD. There is another PCIe4.0 SSD slot available for up to 8TB, these SSD slots are compatible with RAID0 and RAID1, you can store movies, videos, photos, important files easily. What’s more, it also comes with 1x standard PCIex16 slot(PCIe4.0x4) inside.
Where another approach is usually better
- Text generation, document summarization, conversational interfaces, code, images, or speech: use an appropriate generative or specialist model.
- A highly specialized time-series or optimization problem: benchmark a method designed for that problem.
- A single, stable prediction task where a transparent tree model already performs well: compare the simpler model before adopting a foundation-model platform.
The current RPT model family
SAP’s current documentation lists the original models and a newer 1.5 family. The original sap-rpt-1-small is intended for medium-complexity scenarios where latency and throughput matter. sap-rpt-1-large is intended for more complex scenarios where prediction quality and lower error rates are the priority. Documentation also lists sap-rpt-1.5 and sap-rpt-1.5-large; SAP says the 1.5 family can predict multiple target columns together, and results can differ from predicting those targets separately (RPT-1.5 documentation).
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →SAP announced general availability of the original small and large models in Q4 2025 through the generative AI hub in SAP AI Foundation, an open-source version through Hugging Face and GitHub, and a public playground. Verify the current repository, license, hardware requirements, and feature parity before treating the open-source release as equivalent to the managed service.
Documented limits for the original models
| Capability | sap-rpt-1-small |
sap-rpt-1-large |
|---|---|---|
| Maximum context | 2,048 rows | 65,536 rows |
| Recommended context | 500–2,000 rows | 4,000–8,000 rows |
| Maximum columns | 100 | 256 |
| Simultaneous prediction rows | Up to 128 | Up to 128 |
| Simultaneous prediction columns | Up to 10 | Up to 10 |
| Recommended target classes | 256 | 1,024 |
The target-class figures are recommendations for quality, not stated absolute hard limits. The free playground is much smaller—up to 25 rows, four columns, and 50 target classes—so it is suitable for exploration, not production-scale performance testing (SAP architecture guidance).
Rank #3
- Built for Local AI and Advanced Workflows – The BOSGAME M5 AI Mini PC is powered by AMD Ryzen AI Max+ 395 with 16 cores, 32 threads, up to 5.1GHz, 50 TOPS NPU performance and up to 126 TOPS total AI performance. It is designed for local AI inference, private AI assistants, coding, data analysis, virtualization, content creation and demanding multitasking while keeping sensitive data on the device.
- 128GB Unified Memory for Large Models and Creative Projects – M5 includes 128GB LPDDR5X-8000 unified memory, giving the CPU and Radeon 8060S graphics access to a large shared memory pool. This helps support memory-intensive AI workloads, large project files, multiple virtual machines, 3D work, video editing and complex professional applications without the capacity limits of typical 32GB or 64GB mini computers.
- Radeon 8060S Graphics for Creation, Rendering and Gaming – Integrated Radeon 8060S graphics with 40 RDNA 3.5 compute units delivers high-end visual performance without a separate graphics card. Use the M5 creator workstation for 4K video editing, 3D rendering, CAD, AI image workflows, high-resolution media and modern gaming, while maintaining a compact desktop footprint.
- 2TB PCIe 4.0 SSD and Flexible Expansion – A pre-installed 2TB NVMe PCIe 4.0 SSD provides fast access to models, datasets, media libraries and project files. A second M.2 2280 PCIe 4.0 slot allows additional storage expansion, while the SD 4.0 card reader supports efficient photo and video workflows for creators and production teams.
- Professional Connectivity and Four-Display Support – Dual USB4 ports, HDMI 2.1 and DisplayPort 1.4 support up to four displays and resolutions up to 8K@60Hz. WiFi 7, Bluetooth 5.4 and 2.5GbE deliver fast networking for cloud collaboration, NAS access and business deployment. Windows 11 Pro, performance-mode switching, Wake-on-LAN and auto power-on support flexible workstation use.
How an enterprise accesses RPT-1
- Create or use an SAP BTP subaccount.
- Provision SAP AI Core and select the Extended service plan when production generative AI hub access is required.
- Create a configuration under the
foundation-modelsscenario. - Create or deploy the model using executable ID
aicore-sap. - Select the required model name and either an explicit supported version or
latest. - Consume the deployment through inference APIs or the SAP Cloud SDK for AI.
The practical flow is SAP BTP subaccount → AI Core Extended plan → foundation-models scenario → configuration → deployment → aicore-sap → model and version → inference. SAP publishes an RPT regression client example in its Cloud SDK for AI documentation. Do not assume third-party endpoint examples remain current.
For the quickest experiment, use the public playground at rpt.cloud.sap. For a governed prototype or production integration, use SAP AI Core and AI Launchpad.
What SAP’s performance claims prove—and what they do not
SAP says RPT-1 delivers up to 2× better prediction quality than narrow models and 3.5× better prediction quality than LLMs (SAP announcement). Those are vendor claims, not independently established benchmarks in the cited material. A purchasing decision should request the datasets, metrics, baselines, task definitions, confidence intervals, latency, and cost assumptions behind them.
Rank #4
- Speed up your tasks with AI: Unlock new levels of productivity and creativity by upgrading to Intel Core Ultra processors with built-in AI.
- Supports multiple monitors: Connect up to four FHD monitors using DisplayPort and Daisy Chaining*. Or connect two 4K displays using HDMI 2.1 port and DisplayPort.
- Effortless upgrades: The tool-less entry and removable side panel let you quickly access the internal components, making upgrades convenient and stress-free.
- Ready for business: Keep your data secure with a hardware TPM security chip. And when you need to step away from your desk, simply secure your desktop using the built-in lock slot or padlock loop.
- Style meets sustainability: Dell Tower Desktop seamlessly combines elegance with sustainability. Its sleek, modern design, crafted from recycled materials and featuring refined corners, makes it a stylish addition to any home or office.
Cost and platform commitment
No fixed public per-call RPT-1 price is established in the cited SAP material. The public playground is free for initial experimentation, and SAP advertises a 30-day guided generative AI hub trial. Production economics depend on the BTP account and region, AI Core service plan, running deployment resources, model and token usage, request volume, context size, monitoring, integration, governance, and implementation support. SAP describes the Extended plan as resource charges plus model/token usage; use its service-plan documentation and current cost calculator for an actual estimate.
A defensible proof-of-concept plan
- Choose one measurable prediction tied to a real decision.
- Define the target, prediction horizon, intervention, and acceptable error.
- Build a historical dataset with leakage-safe features and representative rows.
- Try the playground to verify the table pattern, not production performance.
- Establish a simple conventional baseline such as a tree-based model.
- Compare small, large, and relevant 1.5 variants across realistic context lengths.
- Measure quality, latency, cost per prediction, calibration, and operational usefulness.
- Review privacy, residency, access control, auditability, and explanation requirements.
- Run a limited pilot with fallback rules or a conventional model for low-confidence cases.
- Monitor drift, missingness, category changes, business outcomes, and model-version changes.
Common failure modes
Deployment is unavailable
Check the AI Core plan, region and quota, foundation-models scenario, executable ID aicore-sap, exact model name, and version. If latest fails, try an explicitly supported version and consult current SAP Help and SAP Notes.
Predictions are poor or unstable
Investigate leakage, class imbalance, unrepresentative examples, inconsistent column meanings, drift, missingness, excessive columns, insufficient rows, and the effect of predicting multiple targets jointly. Compare context lengths and model variants against a time- or operations-realistic holdout set.
The prediction cannot be used
Good statistical performance is not enough if latency misses the decision window, cost is excessive, the output cannot be explained or audited, or the business has no action for the result. Treat workflow fit as part of model selection.
When RPT-1 is a good fit
- Your data is tabular or relational and the target is classification or regression.
- You have many related prediction problems and maintaining a separate pipeline for each is costly.
- You already operate SAP BTP, AI Core, AI Launchpad, or SAP business applications.
- You value governed experimentation and want to reduce task-specific training work.
When to choose something else
- You need text, documents, images, speech, code, or a conversational assistant.
- The target is unclear, labels are unreliable, data is sparse, or leakage dominates.
- You require fully self-hosted infrastructure with no SAP BTP dependency.
- Your workload exceeds documented limits or needs a specialized forecasting, ranking, graph, survival, causal, or optimization method.
- A conventional model already meets quality, cost, interpretability, and latency requirements.
Alternatives to compare
| Category | Strengths | Trade-offs |
|---|---|---|
| Custom ML (XGBoost, LightGBM, scikit-learn, CatBoost) | Control, transparent benchmarking, mature deployment | Separate pipelines, retraining, and MLOps for multiple tasks |
| Managed platforms such as SageMaker, Vertex AI, Azure Machine Learning, DataRobot, and H2O.ai | Broad lifecycle, data, deployment, and governance tooling | More model-selection and platform work; no direct RPT workflow |
| Open-source RPT or other tabular models | Self-hosting and lower vendor dependence | Infrastructure, security, serving, evaluation, licensing, and support responsibilities |
| General-purpose LLM platforms | Strong language capabilities and broad application ecosystems | Not inherently optimized for high-volume structured prediction |
Bottom line
RPT-1 is significant because it brings foundation-model ideas to structured business prediction. It may reduce the number of task-specific model-training projects, especially for SAP-centered organizations with many tabular use cases. It does not remove data engineering, validation, monitoring, governance, specialist modeling, or platform costs. The deciding evidence is performance, cost, control, and business usefulness on your own production-like data—not the “no fine-tuning” slogan or an unqualified vendor benchmark.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
