Apertus-70B-Instruct-2509 — benchmark results

Switzerland's fully open 70B instruct model from ETH Zurich, EPFL and CSCS (September 2025), with transparent training on 15T tokens across 1,800+ languages. Provider: Apertus. Released 2025-09-01. Access: Open.

Unified ELO 1532 ± 8, rank #679 of 1776 rated models, from 294 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EuroEval Dutch Summarization - Wiki Lingua NL36.71Score (%)100
EuroEval Icelandic Summarization - RRN40.02Score (%)100
EuroEval Portuguese Summarization - Publico34.74Score (%)100
EuroEval Catalan Summarization - Dacsa CA39.83Score (%)98.7
EuroEval Czech Summarization - Czech News29.9Score (%)98.7
EuroEval Hungarian Summarization - Hunsum33.96Score (%)98.6
EuroEval Lithuanian Summarization - Lrytas34.41Score (%)98.6
EuroEval English Summarization - CNN Dailymail44.93Score (%)97.3
EuroEval Romanian Summarization - Sumo RO37.51Score (%)97.3
EuroEval Estonian Summarization - ERR News31.94Score (%)96.6
EuroEval Swedish Summarization - Swedn38.69Score (%)96.1
EuroEval Italian Summarization - Ilpost SUM38.7Score (%)96

Interactive version: theaggregate.ai/model?slug=apertus-70b-instruct-2509 · How the rankings work · Data refreshed daily, snapshot 2026-07-22.