aya-expanse-8B — benchmark results
Cohere Aya Expanse 8B multilingual model. Provider: Cohere. Released 2024-10-24. Access: Open.
Unified ELO 1455 ± 7, rank #972 of 1776 rated models, from 322 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| LatamBoard - FLORES Bidirectional | 46.97 | Score (%) | 97 |
| LatamBoard - Translation Score | 48.24 | Score (%) | 97 |
| LA Leaderboard - XNLI Galician | 49.77 | Accuracy (%) | 95.6 |
| EuroEval Dutch NLU - DBRD | 91.72 | Sentiment classification Score (%) | 89.5 |
| WMT25 English-Maasai ESA | 6 | ESA Human Score | 88.2 |
| EuroEval French NLU - Allocine | 95.12 | Sentiment classification Score (%) | 88 |
| MT-Bench PL - Roleplay | 9.25 | Judge Score (0-10) | 87.8 |
| MT-Bench PL - STEM | 9.75 | Judge Score (0-10) | 87.8 |
| EuroEval Spanish NLU - Sentiment Headlines ES | 49.16 | Sentiment classification Score (%) | 87.4 |
| EuroEval English NLU - SST-5 | 68.72 | Sentiment classification Score (%) | 86.5 |
| EuroEval Finnish NLU - Tydiqa FI | 71.14 | Reading comprehension Score (%) | 86.5 |
| MT-Bench PL - Writing | 9.3 | Judge Score (0-10) | 85.7 |
Interactive version: theaggregate.ai/model?slug=aya-expanse-8b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.