Cloudflare adds two European open models to Workers AI
Cloudflare is bringing EuroLLM and Switzerland's Apertus, two open models built by public institutions, to Workers AI, arguing that AI sovereignty depends on being able to choose models rather than wall them off.
Cloudflare announced during its Birthday Week that two European open models are coming to Workers AI. Developers can request access now.
EuroLLM supports 35 languages, including all 24 official EU languages. It was built by a consortium that includes Instituto Superior Técnico, the University of Edinburgh, Université Paris-Saclay, Unbabel, Sorbonne University, Naver Labs and the University of Amsterdam, with funding from Horizon Europe, the European Research Council and EuroHPC. It was trained on the MareNostrum 5 supercomputer. The consortium says it beats similar-sized models on EU multilingual benchmarks and machine translation.
Apertus comes from ETH Zurich, EPFL and the Swiss National Supercomputing Centre (CSCS). It was trained on more than 15 trillion tokens covering more than 1,500 languages, and 40% of that data is non-English. The architecture, weights, training data and methods are all published. Its developers say it was designed with the EU AI Act and GDPR in mind, which means honoring training opt-outs and removing personal data. They also report that it significantly outperforms leading models on rare and regional languages such as Romansh and Swiss German.
Cloudflare added models from India, Japan and Singapore last year. It says hundreds of developers have built on them since, including a voice-guided assistant for Indian government forms that works in 22 languages.
The company is also launching workshops that help government cyber agencies and critical infrastructure operators build AI defenses that work with any model. The first runs in Singapore in October. The post does not give pricing or model sizes.