Apertus-8B-Instruct: benchmark results
Provider: Apertus. Released 2025-09-02. Access: Open.
Aggregate Rating 1477 ± 50, rank #2701 of 10916 rated models, from 17 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Grip on LLMs - HonestCityBench - Non-Text Modality | 38 | Appropriate acknowledgement rate (%; LLM judge) | 88.3 |
| LongPIBench - Resume Screening | 95 | Attack success rate (%; share of 100 long synthetic resumes | 85.7 |
| LongPIBench - Code Review | 87 | Attack success rate (%; share of 100 long synthetic pull req | 78.6 |
| LongPIBench - Paper Review | 92 | Attack success rate (%; share of 100 long synthetic conferen | 71.4 |
| Grip on LLMs - HonestCityBench - Outdated Information | 26 | Appropriate acknowledgement rate (%; LLM judge) | 63.3 |
| LongPIBench - Email Summary | 76 | Attack success rate (%; share of 100 long synthetic email th | 57.1 |
| Grip on LLMs - Dutch Simplification - Amsterdam | 40.78 | SARI (0-100) | 56.7 |
| Grip on LLMs - HonestCityBench - Average | 28 | Appropriate acknowledgement rate (%; mean of five categories | 55 |
| Grip on LLMs - Dutch Simplification - INT Duidelijke Taal | 39.31 | SARI (0-100) | 53.3 |
| Grip on LLMs - Dutch Simplification Average | 40.04 | SARI (0-100; mean of two datasets) | 50 |
| Grip on LLMs - Dutch tinyARC | 66 | Accuracy (%; GP-IRT estimate from 100 tinyBenchmarks items, | 48.3 |
| Grip on LLMs - HonestCityBench - Incorrect Premise | 48 | Appropriate acknowledgement rate (%; LLM judge) | 40 |
Interactive version: theaggregate.ai/model?slug=apertus-8b-instruct · How It Works · Data refreshed daily, snapshot 2026-11-08.