Domyn trains EUROPA model at near one-trillion-parameter scale on Germany's JUPITER supercomputer
Domyn, the Milan-based AI company formerly known as iGenius, says it has trained EUROPA, the European Commission-backed open frontier AI model, at about 952 billion parameters, close to one trillion, on Germany's JUPITER supercomputer, the company said in a press release on November 13. Domyn leads the EUROPA consortium with Fraunhofer IAIS and Fraunhofer IIS.
Instead of a newly built cluster, the team used JUPITER, the publicly funded supercomputer already running in Jülich. Domyn's engineers rebuilt the training stack around PyTorch and TorchTitan, using 384 NVIDIA GH200 GPUs for each copy of the model trained in parallel, and raised training throughput 14-fold, from 90 to about 1,300 tokens per second per GPU. That makes it possible to pre-train a trillion-parameter model in about four months, the company says.
Domyn has published the configuration, engineering methods and measured results in its technical blog and contributed fixes to PyTorch and TorchTitan. It argues the work shows Europe can compete in frontier AI by making better use of supercomputers it has already funded.
We hope you enjoyed this article
Consider subscribing to one of our newsletters like Daily AI Brief.
Also, consider following us on social media:
Daily AI Brief
Daily report covering major AI developments and industry news, with both top stories and complete market updates
Whitepaper
Tensordyne Napier: What If One Rack Could Do the Work of Nine?
Tensordyne
This Tensordyne whitepaper presents Napier, an inference-focused AI processor and rack-scale system based on the company’s TDN Math logarithmic number system. It examines infrastructure requirements for large mixture-of-experts and agentic models, compares major inference architecture approaches, and details the TDN AIP processor, TDN72 pod, TDN Link fabric, and Napier Ultra configuration. The paper reports simulation-based performance, cost, and accuracy-validation results, including Tensordyne’s projected comparison of one Napier rack with a nine-rack Nvidia Rubin plus Groq deployment; the chip is reported as taped out and in fabrication.
Read moreYou may also like
Shopify Cuts AI Serving Costs 90% as NVIDIA Details European Retail Deployments
Nebius Expands AI Builder Program Across Europe With $10 Million
Deepgram Runs Nova-3 on AWS Trainium2 With Up to 70% More Audio per Dollar
AWS Opens Trainium Chips to Existing PyTorch Code
Daisytuner raises €6M to make AI models run faster across chips
Daily AI Brief: the AI news that matters, in your inbox.