The LLM benchmark aggregator.

Granite 4.2 8B

IBM · 8.8B · open weights

Strongest in Knowledge work (#157 of 208), weakest in Agents (#243 of 298). Above par in 2 of 14 scopes.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Core abilities42.4−23.2#172/2011/4

General intelligence39.1−29.9#160/1811/3

Long context26−38.6—/0✱✱

Human preference46.1−22.7#236/3541/1

Agents43.2−23.1#243/2982/6

Knowledge work39.2−31.5#157/2082/2

Tool use48−26—/81✱0/1

Reasoning51−14.4—/195✱0/5

Mathematics54.7−9.1—/152✱0/2

Expert reasoning47.5−18.9—/0✱✱

Coding48.7−17.4—/179✱0/6

Agentic coding48.7−18—/173✱0/5

Code generation44.6−20.7—/20✱0/1

§ 3 · Sources

Where the numbers come from

5 publications, 12 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-11-12

LMArena Text 1290

Artificial Analysis ↗1 measureread 2026-11-12

Artificial Analysis Intelligence Index 5

GDPval-AA ↗1 measureread 2026-11-12

GDPval-AA 499

AA-Briefcase ↗1 measureread 2026-11-12

AA-Briefcase 310

✱✱ IFM — K2 Horizon launch ↗the model’s own publisher8 measuresread 2026-11-03

tau3-Banking 7.6%Terminal-Bench 2.1 18.4%SciCode 30.4%AA-LCR 43.3%SWE-bench Verified 47.7%HMMT Feb 2026 66.5%HLE 9.7%BFCL v4 48.4%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=granite-4-2-8b)](https://publicai.io/model-index/m/granite-4-2-8b)