Gemma-SEA-LION-v5-E2B: benchmark results
AI Singapore's SEA-LION v5 E2B multimodal model (Nov 18, 2026, MIT), post-trained on google/gemma-4-E2B-it (5.12B total with embeddings, 2.3B effective). Provider: AI Singapore. Released 2026-11-18. Access: Open.
Aggregate Rating 1559 ± 36, rank #1438 of 10524 rated models, from 132 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| SEA-HELM (Tamil) - NLU | 83.87 | Normalized Score | 100 |
| SEA-HELM (Tamil) - Question Answering | 71.76 | Normalized Score | 98.6 |
| SEA-HELM (Tamil) - Pragmatics | 74.01 | Normalized Score | 91.9 |
| SEA-HELM (Filipino) - Summarization | 23.3 | Normalized Score | 87.8 |
| SEA-HELM (Burmese) - SEA-Safeguard | 83.87 | Normalized Score | 85.1 |
| SEA-HELM (Indonesian) - Pragmatics | 73.12 | Normalized Score | 79.7 |
| SEA-HELM (Filipino) - NLG | 55.72 | Normalized Score | 78.4 |
| SEA-HELM (Indonesian) - Sentiment Analysis | 80.14 | Normalized Score | 78.4 |
| SEA-HELM (Vietnamese) - Sentiment Analysis | 61.71 | Normalized Score | 77 |
| SEA-HELM (Thai) - SEA-IFEval | 89.96 | Normalized Score | 75.7 |
| SEA-HELM (Burmese) - Translations | 84.62 | Normalized Score | 71.6 |
| SEA-HELM (Vietnamese) - NLU | 69.93 | Normalized Score | 71.6 |
Interactive version: theaggregate.ai/model?slug=gemma-sea-lion-v5-e2b · How It Works · Data refreshed daily, snapshot 2026-11-22.