📊 Full opportunity report: AMÁLIA · The Three Hard Questions. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Portugal’s AMÁLIA, a €5.5M government-funded European Portuguese language model, is now operational. However, critical questions about its openness, native data, and objectives are still unresolved, raising concerns about its strategic value.
Portugal’s €5.5 million AI project, AMÁLIA, is now operational with a publicly accessible base version, but fundamental questions about its openness, native language data, and strategic goals remain unresolved, raising concerns about its future development and policy implications.
The project, involving approximately 60 researchers across Portugal’s leading academic institutions, was announced in December 2024 and completed its base version by September 30, 2025. It is built as a continuation of the EuroLLM model, not trained from scratch, and is currently available to 450,000 academic users via the FCT’s IAedu platform. The model demonstrates strong performance on Portuguese benchmarks, outperforming previous open models and most multilingual models on several tasks.
However, critical questions remain about how open AMÁLIA truly is, given the technical and strategic choices made during development. The model’s training involved only about 5.8 billion tokens from Portuguese web archives, representing roughly 5.5% of its extended pre-training data, and the Portuguese component accounts for approximately 17-18% during supervised fine-tuning. The project’s approach contrasts with Italy’s Minerva, which trained from scratch on Italian data, highlighting different strategic choices.
While the technical performance metrics are promising, experts like Duarte O.Carmo have raised concerns about the model’s openness, native-data sufficiency, and the ultimate goals of the project. The final version is expected in June 2026, but it is not yet clear whether these fundamental questions will be addressed before then.
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.
APRENDA GitLab CI/CD: Implemente DevOps com Deploys Automatizados e Feedback Contínuo (Infraestrutura & Automação Brasil) (Portuguese Edition)
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.
LINKTOR Chemistry Molecular Model Kit (444 Pieces), Student or Teacher Set for Organic and Inorganic Chemistry Learning, Motivate Enthusiasm for Learning and Raising Space Imagination, A Fullerene Set
FOR BASIC TEACHING TO ADVANCED SCIENCE: 444 pieces molecular model kit, including 136 atoms, 158 bonds and 150…
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.
Operations Research and Big Data: IO2015-XVII Congress of Portuguese Association of Operational Research (APDIO) (Studies in Big Data, 15)
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.
Building Generative AI Applications with Open-source Libraries: Practical guide to implementing large language models (English Edition)
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 Portugal’s AI Strategy
The unresolved questions about AMÁLIA’s openness, native-language data, and strategic objectives highlight broader issues facing European sovereign-LLMs. These issues impact national AI policies, data sovereignty, and the continent’s ability to develop competitive, transparent models aligned with local languages and values. The case of AMÁLIA exemplifies the risks of deploying large models without clear answers to these foundational questions, which could influence future investments and regulatory approaches across Europe.
European Sovereign-LLMs and the Structural Challenges
Across Europe, multiple nations are developing large language models with public funding, including Italy’s Minerva, Germany’s Aleph Alpha, France’s Mistral, and others. These efforts are often characterized by strategic choices about data sources, openness, and training methods, but the broader pattern reveals a common set of unresolved questions. Many projects are launched with public investment and transparency commitments, yet lack clear frameworks for openness, native data sufficiency, and goal alignment. The AMÁLIA case underscores this structural challenge, as it is a flagship example of a national effort that still faces fundamental questions about its design and purpose.
Historically, European projects have prioritized multilingual capabilities, but the focus is shifting toward native-language models. The technical approach—whether to build from scratch or adapt existing multilingual models—has significant strategic implications, yet consensus remains elusive. The ongoing debate about these choices influences the development, deployment, and regulation of sovereign models across the continent.
“The three questions about openness, native data, and goals are central to understanding what these models can and should become.”
— Duarte O.Carmo
Unanswered Questions About AMÁLIA’s Future
It remains unclear whether the final version of AMÁLIA, due in June 2026, will address the key questions of openness, native data sufficiency, and strategic goals. Details about future multimodal capabilities, further data collection, or policy frameworks are still emerging, and the research team has not publicly committed to specific resolutions of these issues.
Next Milestones and Policy Implications
The immediate next step is the release of the final version of AMÁLIA in June 2026, which is expected to clarify some of the current uncertainties. In parallel, European policymakers and researchers are likely to scrutinize the model’s development process, data transparency, and strategic direction. Broader discussions about establishing clear standards for openness, native-language data, and goal alignment across European sovereign-LLMs are expected to intensify, shaping future investments and regulatory frameworks.
Key Questions
What are the main concerns about AMÁLIA’s openness?
Experts question whether the model’s technical design and training data truly reflect a fully open approach, given the limited native Portuguese data and the model’s underlying architecture based on existing multilingual models.
Will the final version of AMÁLIA address these questions?
It is not yet clear whether the final release in June 2026 will clarify the model’s openness, native data use, or strategic goals, as the development team has not publicly committed to specific resolutions.
Why are these questions important for Portugal’s AI future?
They determine whether AMÁLIA can serve as a reliable, transparent, and strategically aligned tool for Portuguese language and national AI policy, influencing future investments and governance.
How does AMÁLIA compare with other European LLM efforts?
Unlike some models trained from scratch (e.g., Italy’s Minerva), AMÁLIA builds on a multilingual foundation with limited native data, raising questions about its native-language expertise and openness compared to other projects.
Source: ThorstenMeyerAI.com
