The model will respect European Portuguese and aims for cultural representation, utilizing national supercomputing resources funded by the government.

The development of Amália will span 18 months, with a beta version expected to be released by the end of the first quarter of 2025, leading to a multimodal version that analyzes and generates data in text, image, and video formats.

The implementation of Amália in public administration is anticipated to improve efficiency and accessibility.

The Portuguese government announced the development of Amália, a large language model (LLM) designed to enhance the representation of the Portuguese language in AI, during the Web Summit on November 11, 2023.

Amália will utilize existing synergies from prior AI projects, including EuroLLM and GlórIA, and will be trained using curated data from Arquivo.pt, focusing on Portuguese language, culture, and history.

This initiative, which is set to receive an investment of 5.5 million euros, aims to leverage technology for the benefit of businesses and citizens.

Funding for Amália will be exclusively directed towards public entities involved in its development, ensuring national sovereignty and the preservation of Portuguese language variants.

The model has potential applications in healthcare, such as aiding communication for ALS patients through voice cloning in Portuguese.

All versions of Amália will be available for free and open source, promoting academic and public use while contributing to Portugal's AI innovation ecosystem.

With the launch of Amália, Portugal is enhancing its technological innovation leadership, with the final version expected to be available in 2026.

A specialized monitoring committee will be established to ensure ethical practices and advise on potential applications across various sectors.

Amália is expected to foster innovation among Portuguese startups, creating a foundation for new applications that effectively utilize the Portuguese language.