Llama-3.1-OpenScholar-8B — benchmark results

University of Washington and Ai2's Llama 3.1 8B fine-tune powering OpenScholar, a retrieval-augmented system for citation-backed scientific-literature synthesis. Provider: Meta. Released 2024-11-15. Access: Open.

Unified ELO 1445 ± 23, rank #1026 of 1776 rated models, from 7 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Open LLM Leaderboard - IFEval60.64Score72.5
Open LLM Leaderboard - MATH Level 516.54Score64.8
Open LLM Leaderboard - MuSR11.84Score63.8
Open LLM Leaderboard - BBH32.4Score61.1
Open LLM Leaderboard - MMLU-Pro30.09Score60.5
Open LLM Leaderboard - GPQA4.25Score36.8
FACTS Leaderboard24.15Combined Score (%)31.8

Interactive version: theaggregate.ai/model?slug=llama-3-1-openscholar-8b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.