internlm2-base-20B: benchmark results
InternLM team's internlm2-base-20b foundation checkpoint for further domain adaptation. Its card distinguishes the base release from the separately domain-pretrained InternLM2 and aligned chat checkpoints. Provider: Shanghai AI Lab. Access: Open.
Aggregate Rating 1477 ± 44, rank #2699 of 10916 rated models, from 13 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Chinese LLM - HellaSwag | 66.96 | Accuracy (%) | 82.2 |
| Open Chinese LLM - ARC Challenge | 58.11 | Accuracy (%) | 78.8 |
| Open Chinese LLM - WinoGrande | 65.9 | Accuracy (%) | 75.3 |
| Open Chinese LLM Leaderboard | 59.57 | Official Nonzero-Task Mean (%) | 66 |
| Open Chinese LLM - C-Eval Semantic | 74.85 | Accuracy (%) | 65.1 |
| Open Chinese LLM - CMMLU | 57.61 | Accuracy (%) | 65.1 |
| Open Portuguese LLM - ASSIN2 RTE | 91.13 | Macro F1 (%) | 64 |
| Open Portuguese LLM - ENEM | 63.05 | Accuracy (%) | 54.4 |
| Open Portuguese LLM - BLUEX | 52.57 | Accuracy (%) | 53.2 |
| Open Portuguese LLM - OAB Exams | 42.6 | Accuracy (%) | 52.2 |
| Open Chinese LLM - GSM8K | 45.79 | Accuracy (%) | 50.4 |
| Open Portuguese LLM - FaQuAD NLI | 55.97 | Macro F1 (%) | 38.4 |
Interactive version: theaggregate.ai/model?slug=internlm2-base-20b · How It Works · Data refreshed daily, snapshot 2026-11-08.