granite-3.0-8B-base — benchmark results

IBM's Apache-2.0 Granite 3.0 8B base model (October 2024), pretrained on 12T tokens in two stages as the foundation for the enterprise instruct variant. Provider: IBM. Released 2024-10-21. Access: Open.

Unified ELO 1444 ± 13, rank #1029 of 1776 rated models, from 88 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EuroEval Dutch NLU - SQuAD NL78.32Reading comprehension Score (%)88
EuroEval German NLU - Sb10K57.84Sentiment classification Score (%)86.8
EuroEval Italian NLU - SQuAD IT72.62Reading comprehension Score (%)86.2
EuroEval Portuguese NLU - MultiWikiQA PT75.27Reading comprehension Score (%)85.9
EuroEval German NLU - Germanquad61.9Reading comprehension Score (%)83.1
EuroEval Spanish NLU - MLQA ES65.01Reading comprehension Score (%)82.9
EuroEval Portuguese NLU - SST-2 PT82.81Sentiment classification Score (%)80.7
Open LLM Leaderboard - GPQA10.07Score78.5
EuroEval Dutch65.53Average Score (%)78.2
EuroEval Spanish NLU - Sentiment Headlines ES46.12Sentiment classification Score (%)73.1
EuroEval English NLU - SST-566.89Sentiment classification Score (%)68.3
EuroEval Spanish NLU - ScaLA ES25.47Linguistic acceptability Score (%)65.7

Interactive version: theaggregate.ai/model?slug=granite-3-0-8b-base · How the rankings work · Data refreshed daily, snapshot 2026-07-22.