The Portuguese Government announced last Friday, September 4th, that the Portuguese artificial intelligence model Amália will enter a new phase, acquiring additional capabilities to read, listen and see text, voice and images. The Minister of the Presidency, António Leitão Amaro, indicated at the end of the Council of Ministers that the transition from prototype to public service counters is planned, in order to help serve the people.
For the model training, access to data from public facilities and services such as the National Library, the Cinemateca, archives and museums is being considered. An intention to start a dialogue with media companies to expand access to their most recent archives was also announced, with Portugal intending to be an example in the relationship between the national AI model and the media, respecting concerns about compensation for the use of archives.
Amália – Automatic Multimodal Language Assistant with Artificial Intelligence – is part of the National AI Agenda and was developed by a team composed of Nova University Lisbon, Instituto Superior Técnico, University of Coimbra, University of Porto, University of Minho and the Foundation for Science and Technology. The creation process began with the collection and large-scale processing of European Portuguese data, filtered based on its linguistic relevance and quality, with knowledge collected up to 2023.
Iberian and European supercomputers were used for training, such as the Spanish Mare Nostrum 5, the Portuguese Deucalion and the EuroHPC network. A test version of the first Portuguese language model was completed in the first half of 2025, with the final version presented on July 1st.
