Llama-3.1-Tulu-3-8B-SFT Benchmark Results

Llama-3.1-Tulu-3-8B-SFT: benchmark results

Provider: Meta. Released 2024-11-21. Access: Open.

Unified ELO 1482 ± 20, rank #1255 of 2656 rated models, from 396 benchmark results.

Benchmark	Score	Metric	Percentile
EuroEval Ukrainian NLU - Cross Domain UK Reviews	64.17	Sentiment classification Score (%)	98.5
AbstentionBench - subjective - CoCoNot/Humanizing - F1	98.77	Abstention F1 (%)	97.4
AbstentionBench - subjective - CoCoNot/Humanizing - Recall	97.56	Abstention Recall (%)	97.4
EuroEval Icelandic NLU - NQII	60.51	Reading comprehension Score (%)	94.8
AbstentionBench - answer unknown - CoCoNot/Unsupported - Recall	96.67	Abstention Recall (%)	94.7
AbstentionBench - subjective - KUQ/Controversial - Precision	96.15	Abstention Precision (%)	94.7
EuroEval Swedish NLU - Swerec	79.56	Sentiment classification Score (%)	94.2
EuroEval Slovene NLU - MultiWikiQA SL	70.16	Reading comprehension Score (%)	93.1
AbstentionBench - underspecified intent - CoCoNot/Incomprehensible - F1	97.92	Abstention F1 (%)	92.1
AbstentionBench - underspecified intent - CoCoNot/Incomprehensible - Recall	95.92	Abstention Recall (%)	92.1
EuroEval Swedish NLU - Multi Wiki QA SV	77.41	Reading comprehension Score (%)	89.7
EuroEval Finnish NLU - Scandisent FI	92.38	Sentiment classification Score (%)	89.6

Interactive version: theaggregate.ai/model?slug=llama-3-1-tulu-3-8b-sft · How It Works · Data refreshed daily, snapshot 2026-09-19.