Llama 3.1 70B Instruct: benchmark results

Meta's mid-size open Llama 3.1 instruction-tuned model (70B dense) with a 128K context and multilingual support (July 2024). Provider: Meta. Released 2024-07-23. Access: Open.

Unified ELO 1538 ± 1, rank #490 of 1392 rated models, from 825 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
European LLM Leaderboard - Accuracy70.57Average Accuracy (%)100
HELM SeaHELM - UIT-VSFC73.08Macro F1 score100
HREF48.98Average HREF Score (%)100
HumanLikeness - Meaning-180.22Humanlike Score (%)100
HumanLikeness - Overall66.91Overall Humanlike (%)100
HumanLikeness - Sound-188.51Humanlike Score (%)100
LegalEvalHub - LegalBench75.8Raw Metric Average (%, 162 tasks)100
NVIDIA LLM Robustness - AGIEval Accuracy70.81Accuracy (%)100
NVIDIA LLM Robustness - MMLU-Pro Accuracy64.16Accuracy (%)100
Open LLM Leaderboard - IFEval86.69Score100
RedCode75.54Score (%)100
SCORE Robustness (Accuracy)67.02Average Accuracy (%)100

Interactive version: theaggregate.ai/model?slug=llama-3-1-70b-instruct · How It Works · Data refreshed daily, snapshot 2026-09-05.