Llama 3.1 Tulu3 405B — benchmark results

Provider: Meta. Released 2025-01-30. Access: Open.

Unified ELO 1486 ± 38, rank #902 of 1841 rated models, from 7 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
AA MATH-50077.8Accuracy (%)43.9
Epoch AI - Scicode30.21Score42.4
AA MMLU-Pro71.62Accuracy (%)39.5
AA LiveCodeBench29.1Pass@1 (%)33.5
AA GPQA Diamond51.62Accuracy (%)28.3
Artificial Analysis Intelligence Index8.28Intelligence Index25.9
AA Humanity's Last Exam3.46Accuracy (%)3.8

Interactive version: theaggregate.ai/model?slug=llama-3-1-tulu3-405b · How It Works · Data refreshed daily, snapshot 2026-07-25.