IBM Granite 4.0 Small — benchmark results

Provider: Other. Released 2025-10-02. Access: Open.

Unified ELO 1565 ± 61, rank #645 of 1839 rated models, from 7 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
HELM Safety Anthropic Red Team99.6LM Evaluated Safety score (%)83.1
HELM AIR-Bench71.6Refusal Rate (%)60.5
HELM Safety91.2Mean score (self-reported)50.4
HELM Safety HarmBench71.1LM Evaluated Safety score (%)45.3
HELM Safety BBQ92.5BBQ accuracy (%)43
HELM Safety SimpleSafetyTests98LM Evaluated Safety score (%)40.1
HELM Safety XSTest94.6LM Evaluated Safety score (%)36

Interactive version: theaggregate.ai/model?slug=ibm-granite-4-0-small · How It Works · Data refreshed daily, snapshot 2026-08-05.