📊 Full opportunity report: ALIA. The Spanish answer. on ThorstenMeyerAI.com — validation score, market gap, and execution plan.
TL;DR
Spain’s government-backed ALIA project, a 40-billion-parameter multilingual language model, has been released under open-source license. It aims to serve the Spanish-speaking world and demonstrate Europe’s strategic AI capabilities, but benchmark results reveal operational limitations compared to Llama 2.
Spain has officially launched ALIA, a 40-billion-parameter multilingual language model, marking Europe’s most ambitious publicly funded AI project at scale. Developed by the Barcelona Supercomputing Center and funded with over €240 million, ALIA aims to promote Spanish-language AI adoption and demonstrate Spain’s strategic AI independence. The project’s release under an open-source license underpins its significance for European AI sovereignty and multilingual capabilities.
ALIA, trained on 9.37 trillion tokens across 35 European languages and 92 programming languages, was released on HuggingFace under the Apache License 2.0 in April 2025. It is part of Spain’s broader AI strategy, coordinated by the Secretary of State for Digitalisation and Artificial Intelligence (SEDIA), and developed using MareNostrum 5 supercomputing resources. The project is designed to serve as Spain’s institutional answer to the European sovereign-AI question, emphasizing multilingual coverage with a particular focus on Spanish and co-official languages, as discussed in this analysis.
Benchmark results indicate that ALIA’s performance on standard NLP tasks such as XNLI and SQuAD is below that of Llama 2, with accuracy scores around 51.77% and 81.53%, respectively, compared to Llama 2’s higher scores. This suggests a structural capability gap, aligning with the project’s strategic positioning as a Position 3 model—focused on Spanish-language adoption and operational transparency—rather than a top-performing general-purpose model. The project’s leadership emphasizes its goal is to promote widespread use within the Spanish-speaking world, rather than to achieve the highest benchmark scores globally.
Funding for ALIA includes €90 million allocated for MareNostrum 5 upgrades and €150 million dedicated to integrating ALIA into industry, reflecting a public investment of over €240 million. The project is part of Spain’s broader national AI initiatives, which include earlier projects like AMÁLIA and Minerva, and aims to serve as a structural test case for European AI strategies, balancing multilingual coverage with operational credibility.
ALIA.
The Spanish
answer.
€240M+ Spanish public funding · ALIA-40B + Salamandra family · 9.37T tokens · 35 European languages + 92 programming languages · MareNostrum 5 · Apache 2.0 release. The largest publicly funded European national-AI project by cumulative scope — and the empirical test case for the Position 1 vs Position 3 strategic-positioning argument.
This is the tenth standalone essay in the European sovereign-LLM track and the third Tier 2 expansion piece. ALIA is Spain’s institutional answer — the largest EU member state by GDP not yet documented in the track. The project markets itself as Position 1 + Position 2 simultaneously — “Europe’s first public multilingual foundational model.” The benchmark evidence (ALIA-40B 51.77% XNLI_en vs Llama 2 66%) confirms the structural capability gap from Finding 1 of the synthesis essay. The Position 3 framing — Martorell’s “most widely adopted in the Spanish-speaking world” — is operationally honest. €90M MareNostrum 5 upgrade + €150M company integration = €240M+ cumulative scope. Apache 2.0 open-source release + AESIA validation + co-official languages oversampling. Both can be true at once. The Spanish public discourse would benefit from explicit Position 3 strategic positioning.
Six models. Apache 2.0.
The ALIA family operates as a tiered model portfolio. ALIA-40B is the flagship at 40 billion parameters; the Salamandra family scales down to 7B, 2B and instruct-tuned variants; mRoBERTa provides the foundational multilingual baseline. All released under Apache License 2.0 on April 22, 2025 at the HispanIA 2040 event — “Public Code, Public Money” approach.
multilingual
MN5 LLM
edge
target
instruct
encoder
Natural Language Processing with Transformers, Revised Edition
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Four official. Oversampled by factor of 2.
ALIA’s distinctive multilingual coverage strategy. The four co-official Spanish languages are oversampled by factor of 2 in the training corpus — structurally distinct from Apertus’s broad 1,811-language coverage approach. The strategy targets deep coverage of Spanish co-official languages rather than maximum language breadth.
GPT AI Chat: ChatGPT Chatbot
GPT-4 powered
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
ALIA-40B vs Llama 2. 14-point gap.
The empirical evidence Finding 1 of the synthesis essay needed. ALIA-40B at 40 billion parameters with €240M+ public funding and 8+ months MareNostrum 5 training achieves performance below Llama 2 — a 2023 frontier model released approximately 18 months before ALIA-40B. The capability gap is real and consistent with six of seven prior national-project answers documented in the track.
Hugging Face Transformers Cookbook: Fine-Tuning Open-Source Models for Niche Industry Tasks
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Two pilots. Public administration deployment.
The operational deployment targets that validate the Position 3 + Position 4 framing. Public administration deployment is the structurally credible Position 3 + Position 4 strategic positioning — captive demand from Spanish public institutions where Spanish-language specialization is operationally distinctive.
The work is real across the Spanish ALIA case. €240M+ public funding committed. 40B parameter from-scratch model trained on 9.37 trillion tokens. Salamandra family released under Apache 2.0. AESIA validation aligned with EU AI Act transparency standards. Two pilot applications shipped — Tax Agency chatbot and primary care medicine heart failure diagnosis. The Position 1 framing is operationally misleading. ALIA-40B performance below Llama 2 confirms the structural capability gap. The Position 3 framing is operationally honest — Spanish-speaking world adoption, co-official languages oversampling, public administration deployment. Both can be true at once. The Spanish public discourse would benefit from explicit Position 3 strategic positioning.
AI Engineering: Building Applications with Foundation Models
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Implications for European AI Sovereignty and Spanish Language Models
ALIA represents the largest publicly funded European national AI initiative by scope and scale, highlighting Spain’s strategic effort to establish an independent, multilingual AI infrastructure. Its emphasis on open-source release and validation by AESIA underscores a commitment to transparency and operational credibility, even if benchmark performance lags behind global leaders like Llama 2. The project’s focus on Spanish and co-official languages aims to foster widespread adoption within the Spanish-speaking world, aligning with Spain’s broader goal of positioning itself as a key player in European AI sovereignty.
However, the benchmark results reveal a structural capability gap, suggesting that while ALIA is operationally credible for specific linguistic and regional applications, it may not compete directly with top-tier models on universal NLP benchmarks. This underscores a strategic choice to prioritize language coverage, transparency, and regional adoption over raw performance, which could influence future European AI development and policy decisions.
Spain’s AI Strategy and the European Sovereign-AI Landscape
Spain’s ALIA project is part of a broader European effort to develop sovereign AI capabilities, following earlier initiatives such as Portugal’s AMÁLIA, Italy’s Minerva, and pan-European collaborations like OpenEuroLLM and Mistral. Funded by public investments and aligned with EU policies, these projects aim to reduce dependence on US and Chinese AI giants while fostering regional innovation and multilingual AI solutions.
Developed with MareNostrum 5 supercomputing resources, ALIA’s training process involved 12.875 trillion tokens for Salamandra-7B and Salamandra-2B models, and 9.37 trillion tokens for ALIA-40B. The project’s strategic positioning emphasizes multilingual coverage, open-source transparency, and regional language support, contrasting with other models that aim for global dominance. The focus on Spanish and co-official languages aligns with Spain’s national digital transformation plans and the European Union’s emphasis on digital sovereignty.
Previous European projects have demonstrated varying approaches, with some emphasizing commercial deployment and others prioritizing academic and governmental transparency. ALIA’s open-source release and AESIA validation mark a notable step toward operational transparency and regional adoption, although benchmark results highlight ongoing challenges in matching the performance of leading global models.
“Our goal is not to be the best-performing LLM in the world, but the most widely adopted in the Spanish-speaking world.”
— Josep M. Martorell, ALIA project lead
Benchmark Performance and Strategic Limitations of ALIA
While ALIA’s open-source release and validation suggest operational transparency, its benchmark scores indicate a performance gap compared to models like Llama 2. The exact impact of this gap on regional adoption and practical applications remains to be seen, and further evaluations are ongoing. Additionally, the long-term strategic implications of prioritizing language coverage over benchmark performance are still being assessed by policymakers and developers.
Upcoming Developments and Evaluation of ALIA’s Adoption
Future steps include continued benchmarking, integration of ALIA into Spanish industry and government applications, and monitoring of regional adoption rates. The project team plans to publish ongoing performance assessments and expand multilingual capabilities further. Politicians and stakeholders will likely evaluate whether ALIA can meet operational needs and foster regional AI sovereignty, influencing future funding and development priorities within Spain and Europe.
Key Questions
What is the main purpose of ALIA?
ALIA aims to promote Spanish-language AI adoption, demonstrate European AI sovereignty, and serve as a regional, multilingual foundational model rather than competing for top benchmark scores globally.
How does ALIA compare with other models like Llama 2?
Benchmark results show ALIA’s performance is below Llama 2 on standard NLP tasks, indicating a structural capability gap. Its focus is on regional language coverage and operational transparency.
What are the strategic implications of ALIA’s development?
It highlights Spain’s focus on multilingual coverage, open-source transparency, and regional adoption, emphasizing operational credibility over benchmark performance, which influences European AI sovereignty strategies.
Will ALIA be used commercially?
Currently, the project is primarily aimed at government and industry integration within Spain, with open-source availability encouraging broader regional adoption rather than direct commercial competition.
What are the next steps for ALIA?
Further benchmarking, expanding multilingual capabilities, and integrating ALIA into Spanish industry and government workflows are planned, alongside ongoing performance evaluations.
Source: ThorstenMeyerAI.com
