Models/Granite 4.0 H Small

MDL

Granite 4.0 H Small

IBM

Verified against Artificial Analysis · Jul 13, 2026

Capability

percentile index

Price /M

$0.06 / $0.25

in / out

Context

—

Avg score

22.8

10 benchmarks

AA Intelligence

5.2

index

Speed

412 tok/s

TTFT 8.7s

Released

Sep 22, 2025

About

Granite 4.0 H Small is a open-weights model in the Granite family from IBM. Benchmarked on 10 evals, averaging 22.8.

Benchmark scores · 10

avg 22.8 — higher is better

MMLU-Proartificial-analysis · 2026-07

62.4

GPQA Diamondartificial-analysis · 2026-07

41.6

IFBenchartificial-analysis · 2026-07

31.5

LiveCodeBenchartificial-analysis · 2026-07

25.1

SciCodeartificial-analysis · 2026-07

20.9

τ²-benchartificial-analysis · 2026-07

17.3

AIMEartificial-analysis · 2026-07

13.7

AA Long-Context Reasoningartificial-analysis · 2026-07

Humanity's Last Examartificial-analysis · 2026-07

3.7

Terminal-Bench Hardartificial-analysis · 2026-07

2.3

Details

PublisherIBM

FamilyGranite

Parameters—

License—

WeightsOpen

ReleasedSep 22, 2025

Performance & price

Output speed412 tok/s

Latency (TTFT)8.7s

Price in / out$0.06 / $0.25 per 1M

Blended price$0.107 / 1M

Math index13.7

Median across API providers, via Artificial Analysis. Blended = 3:1 input:output. For reasoning models, latency includes thinking time. Methodology →

History

full ledger →

Capability over time

Append-only ledger — every observed change to this model's numbers.

API providers

Artificial Analysis$0.25/M