Llama 2 70B Base — benchmark results
Meta's 70B pretrained base model from the Llama 2 family (July 2023), trained on 2T tokens under the Llama 2 Community License. Provider: Meta. Released 2023-07-18. Access: Open.
Unified ELO 1405 ± 16, rank #1197 of 1776 rated models, from 263 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| TriviaQA | 87.6 | Accuracy (%) | 100 |
| MMLU-by-task - High School Macroeconomics | 74.1 | Accuracy (%) | 99.9 |
| MMLU-by-task - High School Statistics | 63.43 | Accuracy (%) | 99.8 |
| MMLU-by-task - Philosophy | 78.78 | Accuracy (%) | 99.7 |
| MMLU-by-task - Sociology | 90.05 | Accuracy (%) | 99.7 |
| MMLU-by-task - Human Sexuality | 87.79 | Accuracy (%) | 99.6 |
| MMLU-by-task - Nutrition | 78.1 | Accuracy (%) | 99.6 |
| MMLU-by-task - College Biology | 84.72 | Accuracy (%) | 99.5 |
| MMLU-by-task - Electrical Engineering | 65.52 | Accuracy (%) | 99.4 |
| EuroEval Swedish NLU - Swerec | 80.95 | Sentiment classification Score (%) | 99.3 |
| EuroEval Finnish NLU - Scandisent FI | 93.51 | Sentiment classification Score (%) | 99.2 |
| MMLU-by-task - Professional Medicine | 75 | Accuracy (%) | 99.2 |
Interactive version: theaggregate.ai/model?slug=llama-2-70b-base · How the rankings work · Data refreshed daily, snapshot 2026-07-22.