SloBench - Slovene SuperGLUE: leaderboard

SloBench (CJVT, University of Ljubljana) Slovene SuperGLUE leaderboard: the Slovene translation of six SuperGLUE tasks (BoolQ, CB, COPA, MultiRC, RTE, WSC), scored as the site's average over the tasks, 0-100, higher is better. LLM submissions only; the number of few-shot examples differs by model and task.

Metric: Mean task score (%). Source: slobench.cjvt.si. Saturation forecast: Estimated already saturated. 14 models tracked.

Top models

#ModelScoreOverall rank
1Gemma 3 27B (IT)77.94#528
2Gemma 3 12B (IT)74.61#672
3GaMS3-12B-Instruct73.21#706

No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.

Interactive version: theaggregate.ai/benchmark?slug=slobench-slovene-superglue · How It Works · Data refreshed daily, snapshot 2026-11-05.