IBM Granite 4.0 Small with guardian — benchmark results

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

Unified ELO 1589 ± 68, rank #563 of 1839 rated models, from 7 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
HELM Safety Anthropic Red Team99.9LM Evaluated Safety score (%)96.5
HELM Safety SimpleSafetyTests100LM Evaluated Safety score (%)86
HELM AIR-Bench82.1Refusal Rate (%)77.9
HELM Safety HarmBench86.4LM Evaluated Safety score (%)69.8
HELM Safety91.4Mean score (self-reported)52.5
HELM Safety BBQ90BBQ accuracy (%)32
HELM Safety XSTest80.6LM Evaluated Safety score (%)3.5

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