Inference
EN
IBM

Granite 4.2 3B

Granite 4.2 3B is a reasoning model from IBM in the Granite 4.2 family. 14 benchmarks count toward its index, in 6 categories.

  • Open weights
  • Reasoning
  • Input: text
  • Released 2026-08
RankedEnough results for an index.
IndexAll categories, out of 100.33.2±5.8Median 51.0
CoverageIndex weight that has results.85%
SpeedOutput tokens per second.—Median 67/s
First tokenSeconds to the first output token.—Median 1.3 s
Input / 1MUS dollars per 1M input tokens.—Median $0.60
Output / 1MUS dollars per 1M output tokens.—Median $2.55
ContextMaximum tokens in one request.128KMedian 262K
EloLMArena rating and rank.1293±12 #258Median 1422

Run locally

3.66B parameters
16-bit
8.7 GB
8-bit
5.1 GB
4-bit
3.3 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy fits in 8 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.3 counted33.3194 / 218
CodingCode writing and repair.2 counted34.9190 / 209
ReasoningLogic problems and puzzles.2 counted30.4208 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.4 counted27.8246 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.1 counted58.5—
MathMath problems.2 counted34.3137 / 168

Results

14 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Instruction Following BenchmarkInstructionAbove fitFar above this model's index.74.3%32 / 57
τ³-Bench Tool-Agent-User EvaluationAgentic45.8%18 / 32
American Invitational Mathematics Examination 2025Math78.3%30 / 53
Berkeley Function Calling Leaderboard v4Agentic52.4%16 / 26
LiveCodeBench v6Coding69.7%41 / 59
Critical Physics TasksReasoning0.0%150 / 214
Artificial Analysis Humanity's Last ExamKnowledge6.6%166 / 222
GDPval-AA normalizedAgentic0.0%101 / 133
Massive Multitask Language Understanding ProfessionalKnowledge67.8%60 / 75
Artificial Analysis GPQA DiamondKnowledge55.9%166 / 199
Graduate-Level Google-Proof Q&AKnowledge54.8%119 / 141
Artificial Analysis Long Context ReasoningReasoning24.3%190 / 217
Artificial Analysis SciCodeCoding25.3%108 / 114
Artificial Analysis Omniscience AccuracyKnowledge9.2%163 / 164
Harvard-MIT Mathematics Tournament February 2025Math66.7%15 / 15
Scientific Code BenchmarkCoding24.1%35 / 35
Artificial Analysis Openness IndexKnowledge72.2%—

Sources

BenchLMCC BY-NC 4.0

History

5 releases
Index34.1 → 33.2
23 Oct 202612 Nov 2026

Questions

How good is Granite 4.2 3B?
Its index is 33.2 out of 100. That is place 243 of 283 ranked models.
What is Granite 4.2 3B best at?
Math: 34.3 out of 100, place 137 of 168.
What is the context window of Granite 4.2 3B?
128K tokens.
Is Granite 4.2 3B open source?
Yes. Its weights are published.
When was Granite 4.2 3B released?
August 2026.

Badge

Granite 4.2 3B on Inference360: place 243 of 283, index 33.2
HTML
<a href="https://inference360.ai/models/granite-4-2-3b"><img src="https://inference360.ai/badges/granite-4-2-3b.svg" alt="Granite 4.2 3B on Inference360: place 243 of 283, index 33.2" height="20"></a>
Markdown
[![Granite 4.2 3B on Inference360: place 243 of 283, index 33.2](https://inference360.ai/badges/granite-4-2-3b.svg)](https://inference360.ai/models/granite-4-2-3b)

Compare with

Grok 4.1 Fast33.3Nemotron 3 Nano 30B32.9o1-mini33.5Exaone 4.0 32B32.4Granite 4.2 8B37.6

More from IBM

All models
Granite 4.2 30B41.6Granite 4.2 8B37.6Granite-4.0-H-1B16.3Granite-4.0-H-350M12.7Granite-4.0-350M11.9