internlm2-20B: benchmark results
Shanghai AI Lab's 20B InternLM2 base model, pretrained on over 2.3T tokens of English, Chinese, and code with a 200k context window (January 2024). Provider: Shanghai AI Lab. Released 2024-01-17. Access: Open.
Aggregate Rating 1459 ± 18, rank #943 of 1656 rated models, from 48 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Chinese LLM - WinoGrande | 72.69 | Accuracy (%) | 99.4 |
| Open LLM Leaderboard v1 - WinoGrande | 85.56 | Accuracy (%) (5-shot) | 98.6 |
| Open LLM Leaderboard v1 - MMLU | 67.58 | Accuracy (%) (5-shot) | 89.5 |
| Open Chinese LLM - HellaSwag | 68.28 | Accuracy (%) | 86.7 |
| Open LLM Leaderboard v1 - GSM8K | 67.93 | Accuracy (%) (5-shot) | 84.7 |
| Open Chinese LLM - ARC Challenge | 58.87 | Accuracy (%) | 82 |
| Open Chinese LLM - GSM8K | 57.62 | Accuracy (%) | 80 |
| Open Chinese LLM Leaderboard | 63.91 | Official Nonzero-Task Mean (%) | 78.1 |
| Open Portuguese LLM - BLUEX | 58.55 | Accuracy (%) | 73.6 |
| Science Leaderboard | 43.3 | Average Accuracy (%) | 71.2 |
| Open Chinese LLM - C-Eval Semantic | 80.77 | Accuracy (%) | 69.8 |
| Open Chinese LLM - CMMLU | 60.11 | Accuracy (%) | 69.2 |
Interactive version: theaggregate.ai/model?slug=internlm2-20b · How It Works · Data refreshed daily, snapshot 2026-10-20.