SeaLLMs-v3-7B-Chat: benchmark results
Alibaba DAMO Academy's Southeast Asian chat model built on Qwen2-7B, covering 12 regional languages from Indonesian to Burmese (July 2024). Provider: Alibaba. Released 2024-07-03. Access: Open.
Aggregate Rating 1464 ± 13, rank #915 of 1656 rated models, from 240 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ThaiSafetyBench - Information Hazards | 0.22 | Attack Success Rate (%) | 100 |
| ThaiSafetyBench - Misinformation Harms | 10.63 | Attack Success Rate (%) | 95.7 |
| ThaiSafetyBench - Thai Culture-Related Attacks | 10.64 | Attack Success Rate (%) | 95.7 |
| ThaiSafetyBench - Thai Socio-Cultural Harms | 16.98 | Attack Success Rate (%) | 95.7 |
| Open Korean LLM - Ko-GSM8K (flexible extract) | 60.58 | Exact match, flexible extract (%) | 94.1 |
| ThaiSafetyBench - Malicious Uses | 6.8 | Attack Success Rate (%) | 91.3 |
| Open Arabic LLM - Alghafa Multiple Choice Rating Sentiment NO Neutral Task | 81.06 | Accuracy (%) | 89.5 |
| SeaEval - Cultural Reasoning - SG-Eval v2 Open (Zero-Shot) | 55 | Accuracy (%) | 89.1 |
| Open Portuguese LLM - FaQuAD NLI | 79.92 | Macro F1 (%) | 88.7 |
| Open Arabic LLM - Alghafa Multiple Choice Sentiment Task | 42.44 | Accuracy (%) | 88.6 |
| Open Portuguese LLM - ASSIN2 RTE | 92.84 | Macro F1 (%) | 84.2 |
| Open Portuguese LLM - ENEM | 72.64 | Accuracy (%) | 83.8 |
Interactive version: theaggregate.ai/model?slug=seallms-v3-7b-chat · How It Works · Data refreshed daily, snapshot 2026-10-20.