Apertus on APERTVS

Overview

Wed, 11 Feb 2026 11:11:00 +0100

Apertus is an open source Large Language Model (LLM) developed in Switzerland. This documentation shows you how to get started with the LLM, whether as user, researcher, or advanced contributor: we are maintaining this knowledge base for you, and could ✉️ use your feedback.

This website is running on the Hugo static site engine and Lotus docs theme.

Quickstart

Wed, 11 Feb 2026 11:11:00 +0100

There are several ways to test the Apertus model, the complexity of which ranges from creating an account on your website, to installing software on your computer or cloud machine. We cover some of the easiest ways to plug in here.

Using a cloud provider

A number of supporters of the project are listed on our Get Started page. If you would like to use Apertus with one of them, you need to only visit the home page linked.

F.A.Q.

Wed, 11 Feb 2026 11:11:00 +0100

Is there a question you would like to see answered here? Please drop us a line.

What makes Apertus different from other open models?

Apertus is fully open — offering complete access to training data, code, and alignment principles, not just open weights. This openness supports research, collaboration and informed use, creating a valuable tool for the global AI community. See Swiss AI Initiative.

Can I use Apertus just like other AI systems?

Apertus is not a consumer product: this is foundational infrastructure. Designed to support innovation across research, education, government, and industry — while remaining aligned with Swiss and European values of transparency, neutrality, and accountability. Many organizations are currently piloting solutions with the model, and we will update our showcase as new products based on Apertus become available. See also swissinfo 10.2025.

Apertus paper at ACL 2026

Mon, 06 Jul 2026 10:00:00 +0100

Our technical report presented at a leading conference for AI & NLP

SME Circle #4

Wed, 17 Jun 2026 20:00:00 +0100

Growing momentum around practical AI adoption in Switzerland

Apertus Mini

Mon, 15 Jun 2026 10:00:00 +0100

16 small models to demonstrate distillation and quantization

Llamafile

Thu, 11 Jun 2026 11:11:00 +0100

Llamafiles are open-source AI models packaged as a single file, that runs on your laptop, offline.

Built on top of llama.cpp, this enables offline, cross-platform inference with minimal setup—no Python, CUDA, or complex dependencies required. A Llamafile bundles:

Apertus for Ticino

Tue, 17 Mar 2026 10:00:00 +0100

Fine-tuned model powers in-house AI translation

GPAI Training Transparency

Wed, 04 Mar 2026 10:00:00 +0100

Apertus in research headlights at Trinity College Dublin

SME Circle #3

Tue, 03 Mar 2026 10:00:00 +0100

Building a Sovereign Future for Swiss Business

Evaluation

Wed, 11 Feb 2026 11:11:00 +0100

The Swiss AI Initiative is running experiments and training an LLM on the Alps supercomputer, operational at CSCS since September 2024. About 50% of the capacity was used in the Apertus 1.0 release. Here are some technical details of the total capacity of Alps:

Fine-tuning

Wed, 11 Feb 2026 11:11:00 +0100

Fine-tuning is essential to adapt Apertus to your domain or task, whether you are working on specialized knowledge, improving performance on a certain dataset, or creating a custom application. We have crafted this guide to make the process accessible and flexible.

Our team has prepared these recipes, and in the future we will also provide more references from the community. These are sample configurations for training to ensure you can tailor Apertus to your specific use cases efficiently.

Governance

Wed, 11 Feb 2026 11:11:00 +0100

The core philosophy of Apertus is rooted in transparency, accessibility, and community-driven development. At its foundation is the commitment to open source, with the explicit goal of fostering collaboration, innovation, and the widespread adoption of advanced artificial intelligence technologies in a responsible manner. The open source approach is a foundational element of the project’s identity, reflecting the values of collaboration, inclusivity, and the belief that technology should serve the greater good of society.

Licensing

Wed, 11 Feb 2026 11:11:00 +0100

The open-source nature of Apertus encourages the development of new applications, fosters community engagement, and promotes a culture of transparency and accountability in AI development.

Apache license

As a fully open source large language model, Apertus is released under the Apache license, ensuring that the LLM is free to use, modify, and distribute with minimal restrictions. This allows developers and researchers across the globe to contribute to the project, extend its capabilities, or adapt it for their specific needs without facing legal barriers.

LM Studio

Wed, 11 Feb 2026 11:11:00 +0100

Get LM Studio for free by navigating to the official website: https://lmstudio.ai/ or the open source repositories: https://github.com/lmstudio-ai (MIT license) to run local AI models privately on your computer.

Ollama

Wed, 11 Feb 2026 11:11:00 +0100

Ollama is a tool that lets you run and interact with open source large language models (LLMs) on your local machine.

Models downloaded from the Ollama library can be configured and managed most easily with an elegant chat interface. We are working to make Apertus available here in the future.

Open WebUI

Wed, 11 Feb 2026 11:11:00 +0100

Open WebUI is a self-hosted, ChatGPT-style interface. As long as a serving API is OpenAI-compatible, you can point a local Open WebUI instance at a cloud endpoint and chat with models like Apertus. This is a friendly UI that lets you chat with your models without needing to know the details of the API. It is also a popular choice for institutions and small to medium-sized teams.

Safety

Wed, 11 Feb 2026 11:11:00 +0100

In the development and deployment of LLMs, guardrails are mechanisms used to shape the output of an AI system and ensure appropriate behavior. These controls help manage responses across key domains including topic restrictions, predefined interaction flows, style preferences, structured data extraction, and other content guidelines

In the Apertus LLM, various safety features are being designed to prevent the model from accessing or generating sensitive information, such as personal data, copyrighted content, or toxic material. Guardrails in the training process are based on data filtering, model features like the Goldfish objective, parameter unlearning, and data-influence unlearning to prevent models from accessing or generating sensitive information.

SGlang

Wed, 11 Feb 2026 11:11:00 +0100

SGLang is a high-performance open source serving framework for large language models and multimodal models. It is designed to deliver low-latency and high-throughput inference across a wide range of setups, from a single GPU to large distributed clusters.

Transformers

Wed, 11 Feb 2026 11:11:00 +0100

This document outlines how to use the Transformers library with Apertus.

We are currently working on integrating changes for the Apertus 1.5 release. Please stay tuned for updated instructions here.

vLLM

Wed, 11 Feb 2026 11:11:00 +0100

vLLM is a community-driven, efficient library for deploying large language models (LLMs) for inference and serving. Initially developed by the Sky Computing Lab at UC Berkeley, vLLM has evolved through contributions from academia and industry. While the library itself is user-friendly, deployment can be challenging due to its reliance on NVIDIA libraries and CUDA tools. However, for IT teams using DevOps tools like Kubernetes, vLLM’s multiplatform support and integration capabilities are significant advantages.

Open Source LLM Builders

Mon, 09 Feb 2026 10:00:00 +0100

What It Will Take to Enable Global Collaboration

Tech Report

Wed, 17 Sep 2025 11:11:00 +0100

APERTUS V1 TECHNICAL REPORT

Democratizing Open and Compliant LLMs for Global Language Environments

Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG)

Cite as: arXiv:2509.14233v2

https://doi.org/10.48550/arXiv.2509.14233

Apertus 1.0

Tue, 02 Sep 2025 10:00:00 +0100

A fully open, transparent, multilingual language model

About

Mon, 01 Jan 0001 00:00:00 +0000

About Apertus

Apertus is Switzerland’s first large-scale, fully open, multilingual language model. Developed by researchers at ETH Zurich, EPFL, and the Swiss National Supercomputing Centre (CSCS), it represents a new approach to foundation model development: built by public institutions, designed for the public good.

The name comes from the Latin word for “open” — reflecting the model’s defining characteristic. Unlike commercial models developed behind closed doors, Apertus makes its architecture, weights, training data, and methods fully accessible. Please visit the Frequently Asked Questions and our Documentation section for details. Use the Contact page to share ideas and questions with the Apertus team.

Contact the Apertus team

Mon, 01 Jan 0001 00:00:00 +0000

Contact

Stay on top of our news & developments by 🗞️ subscribing to our newsletter.
For technical enquiries, please use the respective forum and issues on:

Hugging Face GitHub

For media requests, communication, and outreach, email:

For general questions and requests, please write to 📧 llm-requests@swiss-ai.org
or share your feedback on this project and website in the Feedback area.

Documentation

Mon, 01 Jan 0001 00:00:00 +0000

Technical Information

The Frequently Asked Questions cover common issues. See our Research collection for an in-depth look at the architecture, training, data mix, and evaluation results. Deployment & compliance information follows:

Apertus Usage Policy

Terms and conditions for model use

Swiss AI Charter

The internal constitution of Apertus

EU Public Summary

Public summary for EU AI Act compliance

EU Code of Practice

Code of practice documentation

Download the official swiss-ai releases on Hugging Face.

Get Started

Mon, 01 Jan 0001 00:00:00 +0000

Try Apertus

Visit the Quickstart page for information on downloading Apertus models to your own computer. Scroll down to see a list of verified cloud providers. The following applications have been built with Apertus:

Public AI Switzerland

A chat interface you can use for free. Login for access to open models, web search, and more.

ZüriCityGPT OSS

A RAG Chatbot demo developed by Liip to ask questions about city laws with the Apertus model.

Hackathons

Mon, 01 Jan 0001 00:00:00 +0000

Hey there awesome participant!

I’m Oleg, the community manager on the Apertus team. We are a research group commited to advancing the state of the art of Large Language Models with some of the best hardware and smartest people around – and we are doing this as part of the Swiss AI Initiative (swiss-ai.org)

This document is aimed at the participants and organizers of hackathons in Switzerland, some of which we are supporting through our extended team and network of members and partners.

Legal Notice

Mon, 01 Jan 0001 00:00:00 +0000

Legal Notice for Models

Transparency Documentation and Code of Practice

Data Protection and Copyright Requests

For removal requests of personally identifiable information (PII) or of copyrighted content, please contact the respective dataset owners or us directly

Output Filter for PII

Currently no output filter is provided. This may change in the future, so please check back for updates on our Hugging Face organization page and website.

Legal Notice for Websites

The Internet sites apertus-ai.org apertvs.ai apertus.swiss, and all other related domains are the property of Ecole polytechnique fédérale de Lausanne (EPFL).

News

Mon, 01 Jan 0001 00:00:00 +0000

Research

Mon, 01 Jan 0001 00:00:00 +0000

Research News

Papers and technical reports from the Apertus project. This list is continuously expanded: please visit our 📖 Zotero group for other shared literature, and the News area for general announcements.

Apertus: Democratizing Open and Compliant LLMs for Global Language Environments

Main technical report — architecture, training methodology, data pipeline, evaluation.

Can Performant LLMs Be Ethical? Quantifying the Impact of Web Crawling Opt-Outs

Fan, Sabolčec, Ansaripour, Tarun, Jaggi, Bosselut, Schlag

Shows that respecting robots.txt opt-outs causes minimal performance degradation.

Search

Mon, 01 Jan 0001 00:00:00 +0000

Share feedback with the Apertus team

Mon, 01 Jan 0001 00:00:00 +0000

Feedback area

Tell us about your vision for Apertus: which challenges are you facing? Is there any information missing here? We will try to answer any question that interests you about the project. See also our contact page.

We thank all our partners, members, contributors and community for your support!

Nota Bene: we are using Framaforms to process your response, to which separate conditions apply.

Share use cases with the Apertus team

Mon, 01 Jan 0001 00:00:00 +0000

Submissions for Showcase

Tell us about your projects based on Apertus. Please try to include some specifics in the description, such as which size or version of the model, whether your solution is open source, or what kind of audience it is for. If you have unresolved questions or issues, please use our contact page to get in touch.

Nota Bene: we are using Baserow to process your response, to which separate conditions apply. See Privacy Policy

Subscribe to the Apertus Newsletter

Mon, 01 Jan 0001 00:00:00 +0000

Swiss AI Charter

Mon, 01 Jan 0001 00:00:00 +0000

The Swiss AI Charter defines alignment principles for AI systems developed under the Swiss AI Initiative, rooted in Switzerland's constitutional values and democratic traditions.

Version 1.0
August 2025

Preamble

This charter sets forth principles for the alignment of artificial intelligence systems developed under the Swiss AI Initiative. Rooted in Switzerland's constitutional values, democratic traditions, and shared commitment to human dignity, these principles are designed to translate abstract values into concrete alignment criteria for training large language models (LLMs). As AI capabilities advance and our understanding of alignment matures, this charter will adapt through participatory refinement, ensuring that our approach remains both principled and responsive to social and technological change.

Articles

Response Quality — Writing clear, accurate, and useful responses.
Knowledge and Reasoning Standards — Using verified facts and sound reasoning.
Respectful Communication — Treating people with courtesy, fairness, and accessibility.
Preventing Harm — Protecting safety and refusing harmful requests.
Resolving Value Conflicts — Handling trade-offs openly and preserving principles.
Professional Competence Boundaries — Educating without giving licensed advice.
Collective Decision-Making — Supporting fair and constructive group decisions.
Autonomy and Personal Boundaries — Respecting choice, privacy, and clear limits.
Long-term Orientation and Sustainability — Considering long-term impacts and risks.
Human Agency — Keeping humans in control and independent.
AI Identity and Limits — Being clear about what the AI is and is not.

Charter Text

1. Response Quality

The AI should ensure that every response is helpful, harmless, and honest [1.1]. Accuracy, completeness, and usefulness must always take priority, with factual correctness placed above style or polish [1.2]. Each response should fully address the user's question with a level of detail and complexity that matches the scope of the request, keeping explanations concise and proportionate [1.3]. Responses should provide guidance that helps users solve their problems or answer their questions [1.4], while offering clear, actionable steps when guidance or instructions are requested [1.5]. Clarity should be prioritized so that responses are easily understood by the intended audience, favoring simple, accessible, and direct approaches when appropriate for understanding and sound decision-making [1.6].

Apertus on APERTVS

Overview

Quickstart

Using a cloud provider link

F.A.Q.

What makes Apertus different from other open models?

Can I use Apertus just like other AI systems?

Apertus paper at ACL 2026

SME Circle #4

Apertus Mini

Llamafile

Apertus for Ticino

GPAI Training Transparency

SME Circle #3

Evaluation

Fine-tuning

Governance

Licensing

Apache license link

LM Studio

Ollama

Open WebUI

Safety

SGlang

Transformers

vLLM

Open Source LLM Builders

Tech Report

Democratizing Open and Compliant LLMs for Global Language Environments link

Apertus 1.0

About

About Apertus

Contact the Apertus team

Contact

Documentation

Technical Information

Apertus Usage Policy

Swiss AI Charter

EU Public Summary

EU Code of Practice

Get Started

Try Apertus

Public AI Switzerland

ZüriCityGPT OSS

Hackathons

Legal Notice

Legal Notice for Models

Transparency Documentation and Code of Practice

Data Protection and Copyright Requests

Output Filter for PII

Legal Notice for Websites

News

Research

Research News

Apertus: Democratizing Open and Compliant LLMs for Global Language Environments

Can Performant LLMs Be Ethical? Quantifying the Impact of Web Crawling Opt-Outs

Search

Share feedback with the Apertus team

Feedback area

Share use cases with the Apertus team

Submissions for Showcase

Subscribe to the Apertus Newsletter

Swiss AI Charter

Preamble

Articles

Charter Text

1. Response Quality

Using a cloud provider

Apache license

Democratizing Open and Compliant LLMs for Global Language Environments