aya-23-8B — benchmark results

Cohere Aya 23 8B multilingual instruction-tuned model. Provider: Cohere. Released 2024-05-23. Access: Open.

Unified ELO 1371 ± 14, rank #1337 of 1776 rated models, from 155 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open Japanese LLM - SUM12.13Score (%)92.1
Open Japanese LLM - Xlsum JA Bert Score JA F170.72Score (%)90.1
Open Japanese LLM - Xlsum JA Rouge212.11Score (%)90.1
Open Japanese LLM - Xlsum JA RougeLsum26.03Score (%)85.3
Open Japanese LLM - Xlsum JA Rouge130.82Score (%)84.5
EuroEval French NLU - Allocine94.35Sentiment classification Score (%)77.4
Open Persian LLM - ARC Easy80.46Accuracy (%)70
EuroEval Portuguese NLU - HAREM46.27Named entity recognition Score (%)67.5
Open CoT - LogiQA5.11CoT Gain (%)63
EuroEval Dutch NLU - DBRD89.02Sentiment classification Score (%)61.9
Open Persian LLM - ARC Challenge64.43Accuracy (%)60
Open PL LLM - RAG60.76Average RAG Score (%)59.1

Interactive version: theaggregate.ai/model?slug=aya-23-8b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.