Llama 3 8B Cpt Sea Lionv2.1 Instruct — benchmark results

Provider: Meta. Access: Open.

Unified ELO 1400 ± 13, rank #1203 of 1756 rated models, from 40 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
SeaEval - Dialogue - SAMSum (Zero-Shot)30.5Average ROUGE (0-100)87.5
SeaEval - Dialogue - DialogSum (Zero-Shot)25.38Average ROUGE (0-100)78.3
SeaEval - Cultural Reasoning - SG-Eval v1 Cleaned (Zero-Shot)66.18Accuracy (%)73.8
SeaEval - FLORES Translation - Vietnamese-to-English (Zero-Shot)32.83BLEU (0-100)69.6
Sahabat-AI - Indonesian (ID)54.35Score (%)62.5
SeaEval - FLORES Translation - Malay-to-English (Zero-Shot)39.04BLEU (0-100)56.5
SeaEval - FLORES Translation - Indonesian-to-English (Zero-Shot)38.89BLEU (0-100)52.2
SeaEval - Cultural Reasoning - PH-Eval (Zero-Shot)56Accuracy (%)43.8
SeaEval - Cultural Reasoning - SG-Eval (Zero-Shot)66.02Accuracy (%)42.9
SeaEval - Cross-Lingual Consistency - Cross-LogiQA (Zero-Shot)49.92Accuracy (%)40.9
SEA LLM Leaderboard - SeaExam54.1Private Average Score (%)40
Sahabat-AI - Indonesian Average38.29Average Score (%)37.5

Interactive version: aibenchmarks.dev/model?slug=llama-3-8b-cpt-sea-lionv2-1-instruct · How the rankings work · Data refreshed daily, snapshot 2026-07-17.