Granite 4.0 H 1B — benchmark results

IBM's ~1.5B hybrid Mamba-2/attention Nano-series model, Apache-2.0 with a 128K context for edge and on-device use. Provider: IBM. Released 2025-10-01. Access: Open.

Unified ELO 1413 ± 12, rank #1164 of 1776 rated models, from 283 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
LatamBoard - Spanish TeleIA71.43Score (%)86.4
LatamBoard - Spanish MGSM12.8Score (%)81.8
LatamBoard - Spanish Score62.42Score (%)81.8
LatamBoard - Translation Score44.15Score (%)78.8
LatamBoard - FLORES Bidirectional42.71Score (%)75.8
LatamBoard - Spanish OpenBookQA36.4Score (%)75.8
EuroEval Swedish NLU - Swerec77.27Sentiment classification Score (%)74.3
EuroEval Hungarian NLU - HuSST52.81Sentiment classification Score (%)73.5
LatamBoard - Spanish Escola70.37Score (%)72.7
EuroEval Catalan NLU - Guia CAT62.7Sentiment classification Score (%)68.2
EuroEval Finnish NLU - Scandisent FI90.31Sentiment classification Score (%)66.5
EuroEval Dutch NLU - DBRD89.27Sentiment classification Score (%)64.9

Interactive version: theaggregate.ai/model?slug=granite-4-0-h-1b · How the rankings work · Data refreshed daily, snapshot 2026-07-22.