📊 Full opportunity report: AMÁLIA · The Three Hard Questions. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Portugal’s AMÁLIA, a €5.5 million European Portuguese language model, is operational but raises three key questions about openness, native data sufficiency, and optimization goals. These issues have broader implications for Europe’s national AI efforts.
Portugal’s €5.5 million state-funded language model, AMÁLIA, is now operational, with the base version released in September 2025. This development marks a significant step in Portugal’s effort to establish a sovereign European Portuguese large language model, but it also prompts critical questions about its openness, native data use, and strategic aims, which are central to broader European AI sovereignty debates.
AMÁLIA was developed by a consortium of about 60 researchers across Portugal’s leading institutions, including NOVA, IST, and IT, with the project announced in December 2024. The model is based on a continuation of the EuroLLM multilingual foundation, rather than training from scratch, contrasting with Italy’s Minerva approach.
The technical approach involved extended pre-training on European Portuguese data, with approximately 5.8 billion tokens from Portugal’s web archive, Arquivo.pt, comprising about 5.5% of the pre-training corpus. The supervised fine-tuning phase included roughly 17-18% Portuguese data, with no separate native-language pre-training emphasis.
The model outperforms previous open models on Portuguese benchmarks and beats Qwen 3-8B on most Portuguese tasks, though it still trails Qwen on certain benchmarks like ALBA. The final version is due in June 2026, and the project’s strategic questions remain open, especially around how open the model truly is and whether the native data used is sufficient for future needs.
AMÁLIA
The three hard
questions.
Portugal spent €5.5M to build a European Portuguese LLM. The base version is operational, the benchmarks beat Qwen 3-8B on most pt-PT tasks. So why are the most important questions still unanswered?
Last month, Duarte O.Carmo published the sharpest public analysis of AMÁLIA — Portugal’s state-funded European Portuguese large language model. He prefaces his critique with the necessary diplomatic apparatus before doing what almost nobody else in the European-sovereign-LLM discourse has been willing to do publicly: asking hard questions about whether the work, as released, actually does what it set out to do. This piece is a structural extension of his analysis. The AMÁLIA case study exposes three hard questions every national LLM effort needs to answer publicly — and the broader European sovereign-LLM movement has been operating without explicit answers to any of them.
Three questions every national LLM effort needs to answer publicly.
Duarte O.Carmo’s framing maps cleanly onto the structural argument. Each question lands specifically in AMÁLIA — and the broader European sovereign-LLM movement has been operating without explicit answers to any of them.
The three questions form a structural feedback loop. Q3 (optimization target) determines Q2 (data volume needed) which conditions Q1 (openness sufficient for community contribution). The European sovereign-LLM movement collectively benefits from these questions becoming standard methodology disclosure, not exceptional critique.
European Portuguese language learning AI models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
107 billion tokens. 5.8 billion clearly pt-PT.
The structurally tractable question with a structurally surprising answer. For a model whose entire stated purpose is European Portuguese prioritization, the native-language share of extended pre-training is 5.5%. The implications cascade into every other question.
AI model training datasets Portuguese
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Olmo standard. AMÁLIA’s current state.
Allen Institute for AI’s Olmo project defines what “fully open” operationally requires. Olmo doesn’t lead frontier benchmarks. That’s not the point. The point is to be the structural reference for openness. AMÁLIA’s “fully open source” claim should track to the operational standard.
large language model open source tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Four strategic positions. AMÁLIA between two and three.
Approximately €100M+ in publicly disclosed European sovereign-LLM funding across the major initiatives. The structural question every project faces: what is the actual competitive position you’re staking? Four options — none mutually exclusive — but each requiring different commitments.
AI model evaluation benchmarks
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Three standards. For AMÁLIA and the movement.
The structural critique generalizes beyond AMÁLIA. Italy, France, Germany, Switzerland, the OpenEuroLLM consortium, and every subsequent national project benefit from public discourse holding national LLM efforts to operational standards on openness, data accounting, and strategic positioning.
The European sovereign-AI agenda is a serious strategic project that deserves serious public discourse. O.Carmo’s analysis is what serious public discourse looks like. Appropriately diplomatic. Structurally rigorous. Willing to ask the hard questions in public when the public investment justifies it. More of this is needed — across every European sovereign-LLM project, not just AMÁLIA.
Implications for European AI Sovereignty and Policy
The questions raised by AMÁLIA’s development reflect broader concerns about how European nations are building and deploying sovereign AI models. The issues of transparency regarding openness, adequacy of native-language data, and strategic objectives are critical for shaping future policy, funding, and research directions across the continent. As multiple countries pursue similar projects, these questions will influence the legitimacy and effectiveness of European AI independence efforts.
Addressing these questions publicly is essential for accountability, especially given the substantial public investment involved. The way Portugal and other European nations answer these questions could set precedents for how sovereign AI initiatives are evaluated and trusted by the public and industry alike.
European Sovereign LLM Initiatives and Their Structural Challenges
Across Europe, several countries and alliances are developing large language models with national or regional funding, including Italy’s Minerva, Germany’s Aleph Alpha, France’s Mistral, and the OpenEuroLLM consortium. These efforts aim to establish AI sovereignty but face common structural questions about transparency, native data sufficiency, and strategic focus.
Most projects, including AMÁLIA, are not trained from scratch but build upon existing multilingual foundations, raising questions about the openness of these models and the actual native-language data they incorporate. The European sovereign-LLM movement is still grappling with defining clear standards and benchmarks for openness and data adequacy, which are vital for public trust and strategic autonomy.
“AMÁLIA is an impressive piece of work, but it raises fundamental questions about openness and native data that the community must address.”
— Duarte O.Carmo
Unanswered Questions About Model Openness and Data Sufficiency
It remains unclear how open AMÁLIA truly is—whether its architecture and training data will be accessible for independent evaluation, and whether the native Portuguese data used is sufficient for future scalability and robustness. The final version due in June 2026 may address some of these gaps, but current disclosures are limited.
Additionally, the strategic objectives behind the model—whether it is primarily for academic, governmental, or commercial use—are still under discussion, with no definitive public stance yet articulated.
Next Milestones and Public Evaluation of AMÁLIA
The upcoming months will see the release of the final version of AMÁLIA in June 2026, which is expected to include further enhancements and possibly more transparency about data and openness. Simultaneously, public and expert evaluations will likely intensify, scrutinizing whether the model meets its strategic goals and adheres to openness standards.
European policymakers and stakeholders will be watching closely to see if Portugal’s approach can serve as a model or cautionary tale for other national AI projects.
Key Questions
What makes AMÁLIA different from other European language models?
AMÁLIA is based on a continuation of a multilingual foundation, rather than training from scratch, and incorporates Portuguese data from Portugal’s web archive. It is also publicly funded with a specific focus on native language performance and national sovereignty.
Are the openness and data used in AMÁLIA publicly available?
As of now, the final version due in June 2026 is not fully transparent about the openness of its architecture or the extent of native data used. More disclosures are expected at that time.
Why are the three questions about openness, data, and objectives important?
These questions determine the trustworthiness, scalability, and strategic value of national LLM efforts. Addressing them transparently is essential for accountability and for establishing European AI sovereignty.
What impact could AMÁLIA have on European AI policy?
If successfully addressed, AMÁLIA could serve as a benchmark for transparency and strategic clarity in European sovereign AI projects, influencing future funding and development standards.
Source: ThorstenMeyerAI.com
