📊 Full opportunity report: AMÁLIA · The Three Hard Questions. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Portugal’s €5.5 million AMÁLIA language model is now active and surpasses several benchmarks. However, critical questions about transparency, native data sufficiency, and objectives remain unresolved, highlighting broader issues in European sovereign-LLM efforts.
Portugal’s €5.5 million investment in the AMÁLIA large language model has resulted in a functional, Portuguese-specific AI system that outperforms many existing models on key benchmarks. While the model is now publicly available to academic users, critical questions about its transparency, data sources, and strategic objectives remain unanswered, raising concerns about the broader European sovereign-LLM movement.
AMÁLIA, developed through a consortium of approximately 60 researchers across Portugal’s top institutions, was officially launched in October 2025. The model is based on a continuation of the EuroLLM multilingual foundation, with a focus on Portuguese, and is designed primarily for academic use. It has achieved notable benchmark performance, surpassing previous open models and most Portuguese benchmarks, although it still trails Qwen 3-8B on one key test.
Despite these achievements, questions persist about the model’s openness, the sufficiency of native-language data, and its primary optimization goals. The project’s technical approach involves extending an existing multilingual model rather than training from scratch, which contrasts with other European efforts like Italy’s Minerva, which trained from scratch on native data. The training data includes approximately 5.8 billion tokens from Portugal’s web archive, representing about 5.5% of the extended pre-training tokens, with supervised fine-tuning comprising roughly 17-18% of the training data.
While the technical progress is clear, the broader implications for transparency and strategic alignment are less certain. Critics, including Duarte O.Carmo, have raised questions about how open the model truly is, how much native data is enough, and what the model should ultimately optimize for, but these issues have yet to be addressed publicly by the project team.
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.
Portuguese language AI model
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.
large language model for academic use
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.
AI transparency 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.
European language model datasets
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 Sovereign-Language AI Development
The development of AMÁLIA exemplifies the broader challenges faced by European countries in building autonomous, native-language AI models. The project highlights the tension between technical achievements and transparency, especially when public investments are involved. The unresolved questions about openness, data sufficiency, and strategic goals could influence future funding, regulation, and international competitiveness of European AI initiatives.
As other nations pursue similar projects, Portugal’s experience underscores the importance of explicit standards and accountability in sovereign-language model development. The outcome of these efforts will shape Europe’s position in the global AI landscape, affecting policy, industry, and academia.
European Sovereign-LLM Efforts and Portugal’s Role
Over the past year, several European nations have announced or launched large language model initiatives, including Italy’s Minerva, Germany’s Aleph Alpha, and France’s Mistral. These projects are part of a broader push to develop autonomous AI systems that serve national interests and reduce dependence on US or Chinese models. Portugal’s AMÁLIA is notable for its public funding, academic-led approach, and focus on Portuguese language performance. However, the landscape remains fragmented, with questions about standards, openness, and data quality still unresolved across the continent.
The European Union has expressed interest in fostering a sovereign AI ecosystem, but concrete policies and benchmarks are still in development. The case of AMÁLIA illustrates the potential and pitfalls of these efforts, especially regarding transparency and strategic clarity.
“The questions about openness, data sufficiency, and objectives are fundamental and yet remain largely unaddressed in the public discourse.”
— Duarte O.Carmo
Unanswered Questions About AMÁLIA’s Openness and Strategy
It is still unclear how open the AMÁLIA model truly is, especially regarding access to training data and model weights. The specific objectives guiding its development—whether for research, policy, or commercial use—have not been publicly clarified. Additionally, the sufficiency of native Portuguese data remains a contentious point, with critics questioning if the current dataset is enough for long-term robustness and strategic independence.
Further, the final version scheduled for June 2026 may address some of these issues, but no definitive commitments or transparency measures have been announced yet.
Next Milestones and Policy Developments for AMÁLIA
The immediate next step is the release of the final version of AMÁLIA in June 2026, which is expected to include multimodal capabilities. Researchers and policymakers will scrutinize this release for transparency, data access, and alignment with strategic goals. Additionally, Portugal’s government and the consortium are likely to face pressure to clarify their openness policies and data strategies, especially as the European Union considers regulations for sovereign AI models.
Over the next 12-24 months, further benchmarks, transparency reports, and policy debates are anticipated, which will shape the future of Portugal’s and Europe’s sovereign AI efforts.
Key Questions
What makes AMÁLIA different from other European LLMs?
AMÁLIA is a publicly funded, academic-led project focused on Portuguese, using a continuation of a multilingual foundation rather than training from scratch. It emphasizes transparency and national strategic interests.
Is AMÁLIA fully open or accessible to the public?
It is not yet clear how open the model truly is. The project has not publicly disclosed details about data access or model weights, raising questions about its openness.
How much native Portuguese data was used in training?
Approximately 5.8 billion tokens from Portugal’s web archive were used, representing about 5.5% of the training data, with supervised fine-tuning comprising roughly 17-18%.
What are the main concerns critics have about AMÁLIA?
Critics question whether the native-language data is sufficient, how transparent the model is, and what the strategic objectives are—whether for research, policy, or commercial purposes.
What will determine the future success of AMÁLIA?
The final release in June 2026, transparency about data and openness, and how well the model’s objectives align with Portugal’s strategic AI goals will be key factors.
Source: ThorstenMeyerAI.com
