An artificial intelligence model built in Italy outperformed its main foreign competitors in translation testing. This is Lara 3, the system just launched by Translated, a Rome-based firm formed in 1999 by linguist Isabelle Andrieu and computer scientist Marco Trombetti. According to La Stampa, the company submitted 16,800 translations from English into 21 languages to professional translators, who evaluated the texts blindly, without having knowledge of the system used to produce them. Translated's results show that 82% of Lara 3 translations were correct, compared to 69% for OpenAI's GPT-5.6 Sol, 67% for Anthropic's Claude Fable 5, 57% for DeepL, and 53% for Google Translate. The Italian model came out on top in 20 of the 21 language pairs tested. Translated's technique is centered on the creation of specialized models that are intended to address specific needs rather than general requirements. Trombetti believes that this method can open up new chances for Italian and European enterprises competing with huge worldwide technology corporations. "You don't have to be the world's leading AI player to be a leader, but you do have to know exactly what problem you want to address and how to solve it", he said to La Stampa.