Apertus 1.5 released – Latest version of Switzerland's open model

July 24, 2026 – by ETH Zurich, EPFL and CSCS

Trained on the Alps supercomputer at CSCS in Lugano, Apertus is one of the world's largest fully open language models, providing transparency into its training data, code and development process. This allows researchers, public institutions and companies to inspect, adapt and improve the model for their own needs.

Today, July 24 2026, ETH Zurich, EPFL and the Swiss National Supercomputing Centre (CSCS) released the latest version of the model, Apertus 1.5.

A long-term vision for sovereign AI

The release of Apertus 1.5 marks another step towards a long-term goal: building a sovereign AI infrastructure that provides an open alternative to proprietary commercial models. Over the coming years, the Apertus team plans to release regular model updates with steadily expanding capabilities for research, education, public administration and industry.

The project is jointly led by professors Martin Jaggi and Antoine Bosselut from EPFL’s School of Computer and Communication Sciences, and Imanol Schlag, research scientist at ETH Zurich.

“Rather than introducing an entirely new model, Apertus 1.5 incorporates feedback from the first release and adds several new capabilities”, explains Jaggi. Most notably, the model can now understand images and audio alongside text, opening the door to a broader range of applications. It also offers improved reasoning, stronger instruction following and better support for tool use.

“Apertus is not about competing with frontier models from private companies, but about providing trustworthy, transparent AI that organizations big and small can build upon with confidence”, says Schlag.“With every release, this foundation becomes more capable.”

Growing adaptation across sectors

The first release of Apertus has already enabled a growing range of applications. In the Canton of Ticino, it powers an in-house AI translation service, allowing sensitive government documents to be translated without relying on commercial providers. Researchers at EPFL have used Apertus as one of the foundation models for MeditronFO, the world's first fully open framework for building medical large language models.

Apertus is also being used in journalism. The Basel-based online news outlet Bajour runs the locally hosted model in its newsroom to prepare local political news for its daily Basel Briefing newsletter. The team has also developed a search and analysis tool based on Apertus that allows journalists to explore years of cantonal parliament transcripts, trace politicians' positions over time and identify inconsistencies between past statements and later actions, uncovering stories that would otherwise take days of manual research.

"As journalists, we value transparency and openness," says Samuel Hufschmid, editor at Bajour. "With Apertus, we know what we're building on. Its open-source nature and transparency make it a natural fit for our newsroom." The tools developed at Bajour are also being made available to other news organizations through the publishing ecosystem We.Publish.

The release of Apertus 1.5 and the Apertus Mini suite is part of a broader roadmap that will see increasingly capable versions become available over the coming years, providing an open-source foundation for AI in Switzerland and beyond.

Access to Apertus 1.5 via CSCS

CSCS is launching an inference service for Apertus 1.5 for all CSCS users and, by extension, the Swiss academic community. The service provides a dedicated API endpoint for seamless access to large language models, including Apertus, enabling users to develop advanced agentic AI applications.

About Apertus

Apertus is a fully open large language model jointly developed by researchers at ETH Zurich, EPFL and partner institutions within the Swiss AI Initiative. It is trained on the Alps supercomputer at CSCS and released under the Apache 2.0 open-source license, allowing both research and commercial use.

Alongside Apertus 1.5, the team has recently introduced Apertus Mini, a suite of 16 compact models demonstrating how distillation and quantization can deliver efficient, high-quality models for resource-constrained deployments. Together, the flagship and mini models provide a broader range of hardware and deployment scenarios.

By making the model fully open, the Swiss AI Initiative aims to accelerate innovation while supporting transparency, accountability and organisations' governance and compliance efforts. The full model suite is available on Hugging Face and documented on https://www.apertus-ai.org/.