Official CSCS documentation for Alps: platforms, clusters, software, storage, and services.
CSCS
About Empa collaboration with CSCS
Section titled “About Empa collaboration with CSCS”Empa currently has a contract with the Swiss National Supercomputing Centre (CSCS) as the main provider of multiple compute resource allocations.
Every Empa laboratory can request access to compute and data services provided by CSCS. The Scientific IT team at Empa is responsible for managing the collaboration with CSCS. We provide Empa researchers with technical support for compute-related problems and long-term projects.
New CSCS users
Section titled “New CSCS users”To get started, contact us via the contact page or by e-mail to compute@empa.ch. We will provide the necessary instructions to create your CSCS account, set up multi-factor authentication, access Alps, and run your first computations.
To kickstart your project, we will offer you some free compute time to familiarize yourself with the system. Once you are ready to scale up your project, you can apply for a larger allocation of computing resources in three different ways:
-
On-demand resources: you pay for the compute time you use. No commitment is required, and you can use the resources whenever you need them. At the end of the year we will bill your laboratory for the resources used. This is the most expensive option.
-
Pre-allocated resources: a fixed amount of compute time is allocated each quarter. Any unused time will be lost at the end of the period. This option is cheaper than the on-demand resources but requires a commitment from your laboratory. Like the on-demand resources, we will bill your laboratory at the end of the year.
-
Production projects: if you have a project that requires a large amount of compute time, you can apply for a production project. This will be reviewed by a CSCS committee and if approved, you will get a fixed amount of compute resources. To learn more about the application process, please visit the CSCS website. This option will not require any payment from your laboratory, but you must submit a strong proposal.
To learn about the prices and the available resources, please visit the CSCS offering pages.
Platform overview
Section titled “Platform overview”CSCS operates the Alps research infrastructure, a general-purpose compute and data infrastructure hosting several platforms and clusters. Projects are granted access to clusters, which are managed by platforms.
Compute
Section titled “Compute”- HPC Platform (Daint, Eiger): GPU-accelerated computing on Daint and CPU-only computing on Eiger for large-scale simulations.
- Machine Learning Platform (Clariden, Bristen): GPU resources for machine learning, training, and data-intensive workloads. Clariden is the production cluster; Bristen is a best-effort cluster for testing, benchmarking, and data preparation.
- Climate and Weather Platform (Santis): GPU-enabled resources for climate and weather modelling.
Depending on your project and allocation, CSCS also provides storage services (file systems, long-term and object storage) and Kubernetes clusters. Contact Empa Scientific IT to discuss which services are available for your project.
For a complete overview of the available resources, please visit the CSCS website.
LLM inference
Section titled “LLM inference”The LLM Inference API service provides managed inference endpoints that are compatible with the OpenAI and Anthropic APIs, running a curated selection of open-weight large language models, for example Apertus, Kimi, and GLM. CSCS deploys, patches, scales, and operates the serving stack, so no infrastructure management is required from users. Note that private model deployment is not supported.
The service is sovereign and private: your data is processed entirely within CSCS infrastructure in Switzerland, and prompts and responses are not recorded.
The API can be used with familiar tooling, for example coding agents such as Claude Code, OpenCode, and VS Code Copilot, or web interfaces such as OpenWebUI. The currently available models, their context sizes, and pricing are listed on the Inference API UI pricing page. Access to the inference service is granted at the project level: the PI or deputy PI creates an inference resource for the project, and project members create API keys through the Inference API UI.
Support
Section titled “Support”Empa Scientific IT is your first point of contact for general questions about compute, getting access to CSCS, and project setup. See the Compute support page for more details. We are also there to help you with technical problems, for example setting up your software environment or getting your jobs to run on CSCS systems. For step-by-step guides, see our how-tos on building a uenv and building and running a Docker image.
If you have a CSCS-specific technical problem that requires direct support from CSCS, please create a ticket in the CSCS ticketing system. For system status and maintenance windows, check the CSCS status portal.
Resources and links
Section titled “Resources and links”The most important CSCS services for Empa users:
Access your CSCS account information and your allocations.
Request technical help: things not compiling, simulations not running, access to k8s clusters, etc.
Current status of CSCS systems, maintenance windows, and announcements.
Official documentation for the CSCS LLM Inference service and its OpenAI- and Anthropic-compatible API.
Create and manage API keys, view usage, and see model pricing (requires CSCS login).
Current status of the CSCS LLM inference service, including maintenance and model changes.
Web console for the Climate and Weather Platform (requires CSCS login).
Web console for the HPC Platform (requires CSCS login).
Web console for the Machine Learning Platform (requires CSCS login).
Upcoming CSCS events, trainings, and user days.