Portugal’s prime minister, Luís Montenegro, first announced the model during the Web Summit in December 2024, according to Executive Digest.

Amália, which is an acronym for ‘Automatic Multimodal Language Assistant with Artificial Intelligence’, is a language model capable of processing, understanding, and generating natural language text based on vast quantities of data.

According to officials, the purpose of the model is to create a technology that aligns with the Portuguese language and culture, eventually reducing dependency on foreign models and laying the groundwork for its use by Portuguese-speaking countries. The Government said that Amália is the first model developed specifically for Portuguese spoken in Portugal.

Open-source infrastructure

All data, models, and code produced will be made available as “open source”, according to the office of the Deputy Minister for State Reform. Individuals, universities, companies, research centres and public entities will be able to reuse, improve, and develop new applications based on its infrastructure.

Amália will not function as a public chat application, similar to existing commercial platforms such as ChatGPT, Gemini, or Claude, but is intended to be integrated into solutions developed by universities, researchers, companies and public entities.

In public Portuguese administration, the model will be available in areas such as citizen services, education, culture, defence and health.

Pipeline and funding

Projects in the pipeline are an assistant for virtual visits to Portuguese museums and monuments, a lesson planning support tool for teachers, a digital citizen service assistant, and a solution to support critical operations and decisions in the Navy, according to Executive Digest.

Financed through the Recovery and Resilience Plan with an investment of €5.5 million, €2.475 million of that investment was allocated to the Nova University of Lisbon and €1 million to the Instituto Superior Técnico. An additional €375,000 will be allocated to each of the universities of Porto, Minho, and Coimbra, and 900,000 euros will be managed by the Foundation for Science and Technology.

Model development

Amália was based on an existing European technological base, EuroLLM-9B, which was then reinforced by Portuguese teams who developed the model into a multimodal application capable of supporting Portuguese text and images.

The development of the model included refining it with new datasets in Portuguese, increasing its capacity and context, strengthening pre-training with more data in European Portuguese, and developing security, evaluation, and performance optimisation mechanisms, according to the Government.

Behind the project

More than 60 researchers and students from Nova FCT, Instituto Superior Técnico, University of Coimbra, University of Minho, University of Porto and the Foundation for Science and Technology represented the consortium responsible for the development.

Funding for Amália’s next phase is secured through the Recovery and Resilience Plan and the State Agency for Technological Reform until the end of 2027.