internlm-20B: benchmark results
Shanghai AI Lab, SenseTime, CUHK and Fudan's InternLM-20B, a 20B base model pretrained on 2.3T tokens of English, Chinese and code data, using a deeper 60-layer architecture. Provider: Shanghai AI Lab. Released 2023-09-20. Access: Open.
Aggregate Rating 1380 ± 12, rank #1261 of 1635 rated models, from 83 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| MMLU-by-task - Medical Genetics | 72 | Accuracy (%) | 97.8 |
| MMLU-by-task - College Physics | 40.2 | Accuracy (%) | 97.6 |
| MMLU-by-task - US Foreign Policy | 90 | Accuracy (%) | 96.3 |
| MMLU-by-task - High School Statistics | 58.8 | Accuracy (%) | 95.9 |
| MMLU-by-task - College Mathematics | 40 | Accuracy (%) | 95.3 |
| MMLU-by-task - Machine Learning | 49.11 | Accuracy (%) | 93.5 |
| MMLU-by-task - College Medicine | 61.27 | Accuracy (%) | 93.3 |
| MMLU-by-task - High School Computer Science | 71 | Accuracy (%) | 92.9 |
| MMLU-by-task - High School Government And Politics | 90.16 | Accuracy (%) | 92.9 |
| MMLU-by-task - College Computer Science | 53 | Accuracy (%) | 92.5 |
| MMLU-by-task - College Biology | 73.61 | Accuracy (%) | 92.3 |
| MMLU-by-task - Nutrition | 69.61 | Accuracy (%) | 92.3 |
Interactive version: theaggregate.ai/model?slug=internlm-20b · How It Works · Data refreshed daily, snapshot 2026-11-05.