Refresh Models —Deals —Regulatory —Sources —

Claude 3 Sonnet vs IBM Granite 4.0 Small

6 SHARED BENCHMARKS

Across 6 shared benchmarks, Claude 3 Sonnet scores higher on 3 and IBM Granite 4.0 Small on 1, with 2 level. The widest gap is harmbench, where Claude 3 Sonnet scores 95.8 against 86.4.

ANTHROPICVSIBM6 SHARED3–1 HEAD-TO-HEAD

BenchmarkClaude 3 SonnetMARGINIBM Granite 4.0 Small

AIR-Bench 202484.780.4

anthropic_red_team99.899.8

bbq9090

harmbench95.886.4

simple_safety_tests100100

xstest85.880.6

Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.