The state-of-the-art in open-source AI for Swiss legal tasks

Back to Articles

If you need an open-weight LLM for Swiss legal work in German, French, or Italian, there is no single best model in 2026. We evaluated 13 open models on three Swiss legal benchmarks, 55,861 scored samples each, and the top five finish within 2.1 points on our composite score.

The model that wins overall is not the best translator. And the best translator is one of the weakest at multiple-choice law questions. So the right question is not "which open model is best" but "best at what."

This post is for anyone deciding which open model to run on Swiss legal text, and for anyone who wants the current numbers instead of vibes. All results come from a single full run: the raw outputs live in the public Hugging Face bucket, and the run.py entry point reproduces every number below.

TL;DR: Choose GLM 5.2 for translation, Nemotron or a DeepSeek V4 variant for summarization and open-ended legal reasoning, and Gemma 4 31B for local deployment. Gemma scores 54.8 overall and leads the MCQ benchmark; it fits on local hardware while every model above it is a much larger MoE served in the cloud.