LLaMA-65B: benchmark results
Meta's LLaMA 65B, the original (v1) LLaMA base model at 65B parameters, mirrored here in Hugging Face Transformers format under Meta's non-commercial research license. Provider: Meta. Released 2023-02-24. Access: Open.
Aggregate Rating 1379 ± 16, rank #1303 of 1656 rated models, from 108 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| ToolBench - WebShop Short | 41.2 | Task Score | 100 |
| HELM | 90.83 | Mean win rate (imported) | 98.5 |
| HELM Classic - IMDB | 96.2 | Exact Match (%) | 98.5 |
| HELM Classic - NarrativeQA | 75.49 | F1 (%) | 98.5 |
| HELM Classic - NaturalQuestions Closed Book | 43.14 | F1 (%) | 98.5 |
| HELM Classic - WikiFact | 42.08 | Exact Match (%) | 98.5 |
| HELM Classic - MMLU | 58.37 | Exact Match (%) | 97 |
| HELM Classic - bAbI | 67.66 | Exact Match (%) | 95.7 |
| HELM Classic - BoolQ | 87.1 | Exact Match (%) | 92.4 |
| HELM Classic - CivilComments | 65.46 | Exact Match (%) | 92.4 |
| HELM Classic - TruthfulQA | 50.76 | Exact Match (%) | 92.4 |
| HELM Classic - GSM8K | 46.6 | Exact Match (%) | 91.2 |
Interactive version: theaggregate.ai/model?slug=llama-65b · How It Works · Data refreshed daily, snapshot 2026-10-20.