Llama 3.1 8B Cpt Sea Lionv3 Instruct: benchmark results

Provider: Meta. Access: Open.

Unified ELO 1497 ± 1, rank #698 of 1392 rated models, from 38 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Thai LLM - Writing7.45Rating (0-10)87.3
Thai LLM NLG - iapp_squad_seacrowd_qa (ROUGE1)63.43ROUGE185.1
Thai LLM NLG - iapp_squad_seacrowd_qa (ROUGEL)62.8ROUGEL85.1
Thai LLM NLG - iapp_squad_seacrowd_qa (ROUGE2)50.73ROUGE283.6
Thai LLM NLU - wisesight_thai_sentiment_seacrowd_text49.79Accuracy (%)78.3
Thai LLM - Roleplay7.5Rating (0-10)78.2
Thai LLM - Coding7.05Rating (0-10)76.1
Thai LLM NLG - flores200_tha_Thai_eng_Latn_seacrowd_t2t (SacreBLEU)33.3SacreBLEU71.6
Thai LLM NLG - flores200_eng_Latn_tha_Thai_seacrowd_t2t (BLEU)24.95BLEU70.1
Thai LLM - STEM6.55Rating (0-10)69.7
SeaEval - Cultural Reasoning - PH-Eval (Zero-Shot)60Accuracy (%)68.8
Thai LLM NLG - flores200_eng_Latn_tha_Thai_seacrowd_t2t (SacreBLEU)35.39SacreBLEU67.2

Interactive version: theaggregate.ai/model?slug=llama-3-1-8b-cpt-sea-lionv3-instruct · How It Works · Data refreshed daily, snapshot 2026-09-05.