Llama 3 70B: benchmark results
Meta's 70B pretrained base model from the original Llama 3 release, with an 8K context - the foundation for its instruct variant (April 2024). Provider: Meta. Released 2024-04-18. Access: Open.
Aggregate Rating 1544 ± 9, rank #631 of 1656 rated models, from 337 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| HELM Lite - LegalBench - International Citizenship Questions | 70.5 | Exact Match (%) | 100 |
| IFEval Leaderboard | 83.31 | Final Score | 100 |
| Open PL LLM - 8TAGS (multiple choice, 5-shot) | 80.58 | Accuracy (%) | 100 |
| RSPC - Temporal Phase Classification (Few-Shot) | 37.4 | Macro-F1 (%) on classification of the relationship's tempora | 100 |
| HELM Lite - LegalBench - Function of Decision Section | 46.59 | Exact Match (%) | 99.4 |
| Open LLM Leaderboard v1 - GSM8K | 76.88 | Accuracy (%) (5-shot) | 99.4 |
| Open LLM Leaderboard v1 - MMLU | 79.23 | Accuracy (%) (5-shot) | 99.4 |
| Open Japanese LLM - MR | 94.8 | Score (%) | 99.2 |
| HELM NarrativeQA | 79.82 | F1 (%) | 98.9 |
| Open LLM Leaderboard - GPQA | 39.77 | Score | 98.9 |
| EuroEval Finnish NLU - Tydiqa FI | 77.11 | Character F1 (%) | 98.8 |
| EuroEval Finnish NLU | 69.49 | NLU Average Score (%) | 98.4 |
Interactive version: theaggregate.ai/model?slug=llama-3-70b · How It Works · Data refreshed daily, snapshot 2026-10-20.