granite-8B-code-base — benchmark results

IBM's Apache-2.0 8B code base model from the May 2024 Granite Code family, pretrained on ~4T tokens across 116 programming languages. Provider: IBM. Released 2024-04-21. Access: Open.

Unified ELO 1420 ± 36, rank #1131 of 1776 rated models, from 11 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
EuroEval Portuguese NLU - HAREM49.33Named entity recognition Score (%)75.5
ShaderMatch13.96Clone Match Rate (%)61.9
EuroEval Portuguese NLU - MultiWikiQA PT68.63Reading comprehension Score (%)49.2
EuroEval Portuguese NLU48.47NLU Average Score (%)45.8
EuroEval Portuguese NLU - SST-2 PT72.75Sentiment classification Score (%)40.3
EuroEval Portuguese34.71Average Score (%)34.2
DuckDB-NSQL45.3Execution Accuracy (%)30.2
EuroEval Portuguese NLU - ScaLA PT3.17Linguistic acceptability Score (%)30.1
EuroEval Portuguese Knowledge12.94Knowledge Average Score (%)22.8
EuroEval Portuguese Common Sense Reasoning1.46Common Sense Reasoning Average Score (%)21
EuroEval Danish Knowledge5.25Knowledge Average Score (%)16.8

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