Llama-SEA-LION-v3-70B-IT: benchmark results

Provider: AI Singapore. Released 2024-12-18. Access: Open.

Aggregate Rating 1597 ± 21, rank #955 of 10516 rated models, from 152 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
FilBench - Cultural Knowledge76.78Category Score (%)100
SEA-HELM (Thai) - SEA-Safeguard77.46Normalized Score100
SEA-HELM (Vietnamese) - Summarization19.81Normalized Score100
SEA-HELM (Indonesian) - Summarization19.14Normalized Score98.2
SEA-HELM (Tamil) - Summarization14.12Normalized Score98.2
SEA-HELM (Thai) - Summarization25.38Normalized Score98.2
SEA-HELM (Vietnamese) - NLG56.27Normalized Score98.2
FilBench - Classical NLP89.99Category Score (%)97.2
SEA-HELM (Indonesian) - NLG56.02Normalized Score96.5
ThaiSafetyBench - Human-Chatbot Interaction Harms12.39Attack Success Rate (%)95.7
GoTo-AI Leaderboard - IndoMMLU61.48Score (0-100)94.4
GoTo-AI Leaderboard - IndoMMLU - Humanities70.28Score (0-100)94.4

Interactive version: theaggregate.ai/model?slug=llama-sea-lion-v3-70b-it · How It Works · Data refreshed daily, snapshot 2026-11-21.