https://xezpeleta.github.io/en/tags/leaderboard/Recent content in Leaderboard on Xabi EzpeletaHugo -- 0.161.1enCopyright © 2021, Xabier Ezpeleta. License CC BY-SA 4.0.Sat, 23 May 2026 00:00:00 +0000https://xezpeleta.github.io/en/blog/zein-hizkuntza-eredu-da-hobea-euskaraz/Sat, 23 May 2026 00:00:00 +0000https://xezpeleta.github.io/en/blog/zein-hizkuntza-eredu-da-hobea-euskaraz/<p><a href="https://www.hitz.eus/eu/node/340">The Latxa model</a> was an important milestone for artificial intelligence in the Basque language. Until then, we had no open-weight model that handled Basque decently. Since then, several others have been created specifically fine-tuned for Basque: <a href="https://www.orai.eus/eu/arrakasta-kasuak/kimu-zerbitzari-propioetan-instalatzeko-euskarazko-txatbota"><em>Kimu</em></a> (Orai) and the <a href="https://huggingface.co/collections/HiTZ/latxa-vl"><em>Latxa Qwen3 VL</em></a> family of models recently released by HiTZ.</p> <p>At the same time, new top-tier local models have been emerging — <em>Qwen</em>, <em>Gemma</em>, etc. — which can be especially interesting for use in agents with external tools. These models are also improving their Basque language capabilities noticeably. <strong>But do they match the Latxa model?</strong></p>