📊 Full opportunity report: AMÁLIA · The Three Hard Questions. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Portugal’s state-funded AMÁLIA LLM, launched in 2025, demonstrates strong technical performance but raises three critical questions about openness, native data sufficiency, and optimization goals. These issues impact the future of European sovereign AI models.
Portugal’s €5.5 million AMÁLIA large language model, publicly launched in October 2025, has achieved notable technical benchmarks but faces critical, unresolved questions about its openness, native-language data, and primary objectives, according to recent analysis by Duarte O.Carmo.
AMÁLIA is a consortium project involving approximately 60 researchers from Portugal’s leading institutions, including NOVA, IST, and IT. The model, based on a continuation of the EuroLLM multilingual foundation, was completed in September 2025 and is currently available to 450,000 academic users through the FCT platform. It handles Portuguese text and is set to include multimodal capabilities in future versions.
The model’s training involved 107 billion tokens, with roughly 5.8 billion tokens from Portuguese web archives, representing about 5.5% of the total. The model outperforms previous open models on European Portuguese benchmarks and surpasses Qwen 3-8B on most Portuguese tasks, though it still trails Qwen on some specific benchmarks like ALBA.
Despite these achievements, Duarte O.Carmo’s analysis raises three core questions: How open is ‘fully open’ in practice? How much native-language data is enough? And what should be the primary goal of the model—general performance or native-language specialization? These questions are central to evaluating the strategic success and transparency of Portugal’s investment, and they reflect broader issues across European sovereign AI projects.
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.
large language model development kit
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.
multilingual AI training datasets
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 model openness and 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 Portuguese language processing software
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 AI Strategies
The questions raised by AMÁLIA’s development highlight systemic issues in Europe’s approach to sovereign language models. They underscore the need for transparency about openness, clarity on native data sufficiency, and alignment of objectives with national policy goals. How these questions are addressed will influence future investments, model design choices, and Europe’s competitiveness in AI.
European Sovereign LLM Efforts Face Common Challenges
Across Europe, nations like Italy, Germany, France, and Norway are investing in native-language large language models, often with public funds and academic collaborations. These efforts are characterized by a shared structural pattern: models are frequently built on multilingual foundations rather than from scratch, and their openness and native data strategies remain opaque. The case of AMÁLIA exemplifies these broader trends, illustrating both technical progress and persistent strategic uncertainties.
“The three questions—how open is ‘fully open,’ how much native data is enough, and what should we optimize for—are critical to understanding the true value and transparency of AMÁLIA.”
— Duarte O.Carmo
Unresolved Strategic and Technical Questions
It remains unclear how open AMÁLIA truly is, given the lack of transparency about its training data and access policies. The sufficiency of native Portuguese data for achieving comprehensive language understanding is also debated. Additionally, the ultimate goal—whether to prioritize native-language performance or broader multilingual capabilities—has not been explicitly defined by the development team, leaving these questions open for future assessment.
Next Steps in AMÁLIA’s Development and Evaluation
The final version of AMÁLIA is expected in June 2026, which will likely address some of these questions through further testing, data transparency, and policy clarifications. The ongoing evaluation by independent researchers and policymakers will be crucial in determining whether the model’s technical progress translates into strategic value aligned with Portugal’s national AI goals. Broader European projects will also watch these developments closely.
Key Questions
What are the main technical achievements of AMÁLIA so far?
AMÁLIA outperforms previous open models on European Portuguese benchmarks and beats Qwen 3-8B on most Portuguese tasks, demonstrating strong performance in native-language understanding.
Why are the questions about openness and native data important?
They determine how transparent, accessible, and effective the model will be, influencing trust, policy decisions, and future AI development strategies at a national and European level.
What are the risks of not addressing these questions?
Failing to clarify openness and native data sufficiency could lead to less transparent models, reduced public trust, and strategies that do not fully serve national language needs or AI sovereignty goals.
Will the final version of AMÁLIA resolve these questions?
It is uncertain. The final version due in June 2026 may clarify some issues, but the structural questions about openness, native data, and goals are broader and may require ongoing policy and technical debate.
Source: ThorstenMeyerAI.com
