Which is better, Granite 4.0 1B or LFM2.5-1.2B-Instruct?

Refresh Models —Deals —Regulatory —Sources —

Granite 4.0 1B vs LFM2.5-1.2B-Instruct

30 SHARED BENCHMARKS

Across 30 shared benchmarks, Granite 4.0 1B scores higher on 9 and LFM2.5-1.2B-Instruct on 18, with 3 level. The widest gap is AIME25 no tools, where LFM2.5-1.2B-Instruct scores 14 against 3.3.

IBMVSLIQUID AI30 SHARED9–18 HEAD-TO-HEAD

BenchmarkGranite 4.0 1BMARGINLFM2.5-1.2B-Instruct

AA Agentic Index7.63.6

AA Intelligence2.12.3

AA-LCR60

AA-Omniscience-81.6-72.1

AIME25 no tools3.314

Artificial Analysis Coding Index2.90.8

Avg.34.150.1

BFCL v352.449.1

critpt00

Domain Avg38.539.7

gdpval00

GPQA Diamond28.138.9

GSM8K73.464.5

HLE5.16.7

IFBench2147.3

ifeval79.686.2

J-BFCLv350.646.3

J-GSM8K42.850.2

J-MATH50025.450

JGPQA24.431.7

JHumanEval+51.228.7

JMMLU33.947.6

JMMLU-ProX15.331.4

math_500_em44.863.2

MMLU-Pro33.544.4

OmniScience Accuracy6.27

OmniScience Non-Hallucination6.514.9

scicode8.72.3

Terminal-Bench Hard00

τ²-Bench Telecom (AA run)22.810.8

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.