Refresh Models —Deals —Regulatory —Sources —

Claude Haiku 4.5 vs IBM Granite 4.0 Small

6 SHARED BENCHMARKS

Across 6 shared benchmarks, Claude Haiku 4.5 scores higher on 4 and IBM Granite 4.0 Small on 2. The widest gap is AIR-Bench 2024, where Claude Haiku 4.5 scores 93.2 against 80.4.

ANTHROPICVSIBM6 SHARED4–2 HEAD-TO-HEAD

BenchmarkClaude Haiku 4.5MARGINIBM Granite 4.0 Small

AIR-Bench 202493.280.4

anthropic_red_team96.999.8

bbq92.890

harmbench95.986.4

simple_safety_tests98.8100

xstest93.280.6

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.