Ctrl K
Back to models

IBM

Granite 4.2 3B

  • Released
  • Open weights
  • Apache 2.0
Compare
Intelligence Index

Reasoning setting

9.1#75 of 83measured models
Measured · Artificial Analysis v4.3.2
Context window
131.07K tok

Performance profile

ReasoningRank #91 of 107
55.9%
GPQA Diamond
CodingRank #85 of 94
25.3%
SciCode
Cost efficiencyRank #1 of 67
1,512.5pts / $
Intelligence ÷ eval cost

Capabilities

Context window

Modalities

Measured setting

Reasoning trade-offs

Source (opens in a new tab)
Intelligence Index v4.3.2
Granite 4.2 3B
Cost per task (USD) · log scale
Granite 4.2 3BReasoning
Intelligence Index
9.1
Cost / task
$0.006
Evaluation time
1.3m

Pricing & access

No API offering listed in our tracked sources.

Run locally

GGUF
Context
Estimated memory

~4.7 GiB

RAM + VRAM

Estimate assumptions

One text chat at the selected context, with a 16-bit cache and app allowance. Operating system and image processing are extra; optimized caches can use less.

16 GiB RAMwithout GPU offload

Memory budgetOne text chat
Weights
2.1 GiB
Context cache
0.6 GiB16-bit cache
App allowance
2 GiB
Download
2.2 GB
Disk space
~8 GB
Available software
Ollama (opens in a new tab)LM Studio (opens in a new tab)Jan (opens in a new tab)KoboldCpp (opens in a new tab)llama.cpp (opens in a new tab)
Get GGUF (opens in a new tab)
Set up in Ollama · Q4_K_M · 8K context

Save the following as Modelfile. The context matches the estimate above.

FROM hf.co/lmstudio-community/granite-4.2-3b-GGUF:Q4_K_M
PARAMETER num_ctx 8192

Then run these commands in your terminal:

ollama create llmcurrent-local -f Modelfile
ollama run llmcurrent-local

Ollama import instructions (opens in a new tab) · For LM Studio or Jan, open the app’s model search, enter lmstudio-community/granite-4.2-3b-GGUF, choose Q4_K_M, and set context to 8,192 tokens in the app’s model settings. Software links above open each app’s setup guide.

vLLM setup (opens in a new tab)

Benchmarks

Full leaderboardSource (opens in a new tab)

Humanity’s Last Exam

Frontier reasoning

Granite 4.2 3B · Reasoning#102 of 126

  1. 1Claude Opus 5.5Max61.4%
  2. 2Claude Fable 5.1Max59.1%
  3. 3Gemini 4 ArgonHigh57.1%
  5. 99Qwen3.5 Omni FlashStandard7.6%
  6. 100Qwen3 Omni 30B A3B (Reasoning)Reasoning7.5%
  7. 101Celeris-1Standard6.8%
  8. 102Granite 4.2 3BReasoning6.6%
  9. 103LFM2.5-2.6BReasoning6.2%
  10. 104Llama 3.2 Instruct 11B (Vision)Standard5.5%
  11. 105NVIDIA Nemotron Nano 12B v2 VLReasoning5.5%

Peers use their best measured result on this benchmark.

Alternatives

Each model at its highest-scoring measured setting · Intelligence Index v4.3.2 and its evaluation cost per task.