Spain’s Instituto Cervantes has published a report counting 19,360 Spanish-labelled open models on Hugging Face through June. Spanish ranks fourth on this measure, with stronger representation in translation than image and video generation. Instituto Cervantes calls for more varied language resources and cooperation between Spain and Latin America.
Spain’s Instituto Cervantes, the institution promoting Spanish language and culture, published its new artificial intelligence report on November 6. Its count reaches 19,360 Spanish-labelled open models on Hugging Face, a platform for sharing models and data, through June. Spanish ranks fourth by this measure, ahead of German and Portuguese. The labels describe the open ecosystem rather than the language proficiency of individual systems.[1], [2]
Spanish represents more than 10 percent in the platform’s translation field, while its presence in image and video generation is smaller. Instituto Cervantes also examines language resources beyond models. In the Common Crawl web collection analyzed, Spanish is the main language of 4.68 percent of identified pages, close to French and above Chinese and Portuguese. Commercial models’ internal language proportions are outside what these open indicators establish.[1]
Instituto Cervantes recommends expanding spoken and audiovisual resources and better representing Spanish varieties used in the Americas. It calls for evaluation tests designed from Spanish, stronger data documentation and stable funding and infrastructure. Organized collections of text and speech help document linguistic variety. Institutions in Spain and Latin America are urged to connect their resources and expertise to support an open infrastructure reflecting that diversity.[1]
