29-01-2025

Almawave, today presented at the “Auditorium della Tecnica Confindustria” in Rome “Velvet”: a family of multilingual generative Artificial Intelligence models. The models – Velvet 14B and Velvet 2B – are foundational and Instruct version Large Language Models (LLMs) which have been fully developed by the company in Italy on its own architecture, and trained on the Leonardo supercomputer managed by Cineca. The models are released in Open Source mode. The Velvet technology models for Enterprises embody the extensive expertise gained by Almawave and its personnel over more than 15 years of R&D through hundreds of AI-related projects in sectors such as Healthcare, Social Security, Justice, Security, Mobility, Finance and the Public Sector. This AI has been carefully designed within the European framework to be sustainable. It is therefore energy-saving, while also highly-performing when applied to real-use cases.

Valeria Sandei, Chief Executive Officer of Almawave, stated: “Velvet emerges from a strategic decision to confidently invest in a realm of technology with vast potential, building an Artificial Intelligence designed to simultaneously condense high-performance, consumption-efficiency and great agility to perform targeted tasks in the vertical sectors in which we operate. Our vision is not to merely adapt to the European framework, but to consider it a key value during construction. This development is the result of our more than 10-year journey in language technologies, which has always focused on technical expertise in AI and which today – and increasingly in the future – can make all the difference”.