>
A
Model Details
Domain:
Model Access:
Open weights (unrestricted)
Introduction
Researchers at EPFL, ETH Zurich and CSCS have developed Apertus, a fully open Large Language Model (LLM) – one of the largest of its kind. As a foundational technology, Apertus enables innovation and strengthens AI expertise across research, society and industry by allowing others to build upon it. Apertus is currently available through strategic partner Swisscom, the AI platform Hugging Face, and the Public AI network.
Benchmarking
FLOPs
6.74e+24Notes: 6.74 · 10^24 FLOPs - reported 6 FLOP / parameter / token * 70 * 10^9 parameters * 15 * 10^12 tokens = 6.3e+24 FLOP 989500000000000 FLOP / GPU / sec [GH200, bf16 reported] * 6*10^6 GPU-hours * 3600 sec / hour * 0.3 [assumed utilization] = 6.41196e+24 FLOP
Training
Training Code Accessibility
Apache 2.0
https://huggingface.co/swiss-ai/Apertus-8B-Instruct-2509
Apache 2.0
https://github.com/swiss-ai/pretrain-codeHardware
NVIDIA GH200Hardware Quantity
4096Size Notes: "Trained on 15 trillion tokens across more than 1,000 languages – 40% of the data is non-English"
Parameters
Parameters
70000000000Notes: 70B