Post by ETH AI Center

The Apertus family is growing. Meet Apertus Mini: a new collection of 16 compact language models. Developed by the Apertus team and supervised through the ELLIS program by Dan Alistarh (Institute of Science and Technology Austria) and Martin Jaggi (EPFL), Apertus Mini uses distillation and quantization to reduce computational requirements while retaining many of the capabilities of larger models. Training was made possible through the CSCS infrastructure, supported by Swiss AI Initiative grants. The result? Faster, more efficient AI that can run under tighter hardware and energy constraints, opening new possibilities for local deployment and broader access to AI. Chat with the Apertus Mini models directly in your browser, with a space contributed by 🤷 🤷 Nico Martin at Hugging Face: https://lnkd.in/e34xnzry Read more about Apertus Mini and subscribe to the Apertus newsletter to stay up to date with the latest developments: https://lnkd.in/eciaMMcU   ETH Zürich ETH AI Center, EPFL AI Center, CSCS, ELLIS - European Laboratory for Learning and Intelligent Systems, Institute of Science and Technology Austria, Imanol Schlag, Andrei Panferov, Davit Melikidze Video by 🤷 Nico Martin

Video Content