ERNIE 4.5 21B A3B: benchmark results
Provider: Baidu. Released 2025-06-28. Access: Open.
Aggregate Rating 1465 ± 28, rank #912 of 1656 rated models, from 42 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CASTLE Student Safety (English) - Academic Integrity and Competence | 1.82 | Average Safety Score (1-5) on CASTLE's English scenarios, th | 88.2 |
| CASTLE Student Safety (English) | 1.87 | Average Safety Score (1-5) on CASTLE's English scenarios, al | 82.4 |
| CASTLE Student Safety (English) - Content and Information Bias | 1.73 | Average Safety Score (1-5) on CASTLE's English scenarios, th | 82.4 |
| CASTLE Student Safety (English) - Psychological and Emotional Health | 2.08 | Average Safety Score (1-5) on CASTLE's English scenarios, th | 82.4 |
| CASTLE Student Safety (English) - Learning Dependence and Cognition | 1.76 | Average Safety Score (1-5) on CASTLE's English scenarios, th | 79.4 |
| Persian Fluency Bench - Conciseness | 25.3 | Judge Score (%, judge google/gemini-2.5-flash) | 60 |
| CASTLE Student Safety (Chinese) - Academic Integrity and Competence | 1.92 | Average Safety Score (1-5) on CASTLE's Chinese scenarios, th | 47.1 |
| CASTLE Student Safety (Chinese) - Excessive Seeking Simplification | 1.83 | Average Safety Score (1-5) on CASTLE's Chinese scenarios, th | 47.1 |
| CASTLE Student Safety (Chinese) - Loss of Independent Judgment | 1.89 | Average Safety Score (1-5) on CASTLE's Chinese scenarios, th | 47.1 |
| Persian Fluency Bench | 27.6 | Judge Score (%, judge google/gemini-2.5-flash) | 46.7 |
| Persian Fluency Bench - Context Retention | 59.4 | Judge Score (%, judge google/gemini-2.5-flash) | 46.7 |
| Persian Fluency Bench - Grammar | 30 | Judge Score (%, judge google/gemini-2.5-flash) | 46.7 |
Interactive version: theaggregate.ai/model?slug=ernie-4-5-21b-a3b · How It Works · Data refreshed daily, snapshot 2026-10-20.