[whatmodel.sh](../../README.md) / Granite 4.0 H Micro 3B
# Granite 4.0 H Micro 3B
IBM · released 2025-09 · 3.19B params · apache-2.0
**score 47.6** ±0.9 · #415 of 732 · from 24 benchmark results across 3 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks Granite 4.0 H Micro 3B among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ●·▄········█ #4 of 4* 1.3% 1
coding ●···▄······▄ #4 of 4 32% 7
chat & writing ▁▁▁▃▆▄▆●▆▇▇▃ #212 of 402* 4.8% 2
math ▁▁▃▅●▆▄█▄▂▁▁ #271 of 431 25% 5
agent tasks · #4 of 4 · 1.3% of its score
Trails [Granite 4.0 Micro 3B](granite-4-0-micro.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
BFCL v3 (maker) 57.6% ▁▁▁▁▁▃▅●▆▅▄▁ beats 84/171 medium · avg 60% 1.3% blunt · at its level
coding · #4 of 4 · 32% of its score
Trails [Granite 4.0 Micro 3B](granite-4-0-micro.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
HumanEval (maker) 81.0% ▁▁▁▁▂▂▆▆█●▂▁ beats 177/205 easy · avg 84% 10% sharp
HumanEval+ (maker) 75.0% ▁·▁▁▂▃▄▄●▅▄▃ beats 96/145 easy · avg 79% 8.5% sharp · at its level
MBPP (maker) 73.0% ▁▁▁▄▅█▆▅●▁▂▁ beats 74/86 easy · avg 78% 3.9% at its level
MultiPL-E (maker) 49.5% ▃··▄·●▅▇▄█▃▄ beats 6/31 medium · avg 56% 3.0% at its level
BigCodeBench (maker) 37.9% ▅····▃·█▃●█▃ beats 8/12 medium · avg 41% 3.0% at its level
MBPP+ (maker) 64.0% ▁··▁▁▃▄●▆▄▁▁ beats 59/120 easy · avg 68% 2.7% at its level
CRUXEval-O (maker) 41.3% ▂▂█●▅▂▁·▁·▁▁ beats 30/51 medium · avg 46% 1.6% blunt · at its level
chat & writing · #212 of 402 · 4.8% of its score
Beats [Gemma 2 9B IT SimPO](gemma-2-9b-simpo.md), [Granite 4.2 8B](granite-4-2-8b.md), [Mistral Small 4 (119B-A6.5B)](mistral-small-4.md) here, though they rank higher overall. Trails [LFM2.5 1.2B Instruct](lfm2-5-1-2b-instruct.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
IFEval instruction-level strict (… 86.9% ▄▄····▄▄█●·█ beats 6/8 very easy · avg 88% 2.4% typical
IFEval prompt-level strict (maker) 81.7% ▁▁··▂··▂●█▃▃ beats 10/26 easy · avg 80% 2.4% typical
math · #271 of 431 · 25% of its score
Beats [GPT-4 Turbo (2024-04-09)](gpt-4-turbo-2024-04-09.md), [Molmo 2 8B](molmo-2-8b.md), [Llama 3.3 70B Instruct](llama-3-3-70b.md) here, though they rank higher overall. Trails [DeepSeek R1 Distill Qwen 1.5B](deepseek-r1-distill-qwen-1-5b.md), [Gemma 7B (base)](gemma-7b-base.md), [CodeGemma 7B (v1.0)](codegemma-7b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
GSM8K (maker) 81.3% ▂▂▂▂▃▄▅▇█●▆▄ beats 143/206 very easy · avg 86% 8.9% sharp
Minerva MATH (maker) 66.4% ▁·▃▁▁▂▁█▄▁●▃ beats 22/27 easy · avg 68% 5.4% at its level
DeepMind Mathematics (maker) 43.8% ▅····▅·●▃▃·▅ beats 4/10 medium · avg 53% 5.0% at its level
GSM-Symbolic (maker) 77.5% ▃▃····▃·▅●█▃ beats 5/10 easy · avg 82% 4.2% typical
MGSM, 5 languages (en, es, fr, ja… 44.5% █·▄▄█●···█·▄ beats 6/10 medium · avg 44% 1.7% at its level
other tests (9)
test result vs the others in its run difficulty counts why
BIG-Bench Hard (maker) 69.4% ▁▁▁▁▂▄▅▆●▃▃▁ beats 112/155 easy · avg 74% 7.1% sharp · at its level
MMLU (maker) 67.4% ▁▁▁▁▁▂▂▅█●▆▄ beats 120/214 medium · avg 73% 6.7% at its level
GPQA (subset unspecified) (maker) 32.1% ▂▁▂▂▄●▆▃▅█▃▃ beats 27/106 very hard · avg 35% 6.5% sharp
MMLU-Pro (maker) 43.5% ▁▁·▁▁▁●▂▄▅█▁ beats 34/188 medium · avg 50% 4.8% at its level
INCLUDE (maker) 50.5% ▂····▂●▄▄▄█▄ beats 5/36 medium · avg 57% 4.0% at its level
MMMLU (maker) 55.2% ▂▁▁▂▃▂●▄▅██▃ beats 12/46 medium · avg 62% 3.6% at its level
EuroEval HellaSwag (English), tes… 67.7% ▃·▁▁▂▁▂▅●▇▄▁ beats 43/100 medium · avg 67% 1.6% sharp · at its level · noisy
EuroEval Life in the UK (English … 81.6% ▁▁▁▁▁▁▁▂▄●▄▃ beats 73/146 easy · avg 82% 1.1% at its level · noisy
Vectara hallucination leaderboard… 91.6% ▁▁▂▃●▄█▆▃▂▂▁ beats 32/140 very easy · avg 94% 0.6% blunt · maxed out · noisy
source results share of the score
ibm-granite/granite-4.0-h-small model c… 21 96% ▇▇▇▇▇▇▇▇▇▇▇▇
EuroEval leaderboards 2 3% ▇
Vectara hallucination leaderboard 1 1% ▇
prior: pull toward the middle 0% ▇
weighted average = 47.6
Licences: ibm-granite/granite-4.0-h-small model card ([Apache-2.0](../../licenses/ibm-granite--granite-4.0-h-small.txt)) · EuroEval leaderboards ([MIT](https://github.com/EuroEval/leaderboards/blob/1a77f1d2f8ba1ba05ab73287fb090d76bf41b2e0/LICENSE)) · Vectara hallucination leaderboard ([Apache-2.0](https://www.apache.org/licenses/LICENSE-2.0))
## How to run it
### llama.cpp (recommended)
1. Install llama.cpp. Homebrew and winget installs name the commands llama-cli and llama-server instead. [Other ways to install](https://github.com/ggml-org/llama.cpp/blob/master/docs/install.md)
```sh
# macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
# Windows
irm https://llama.app/install.ps1 | iex
```
2. Run Granite 4.0 H Micro 3B (Q4_K_M, needs ~1.94 GB, from [huggingface.co/ibm-granite](https://huggingface.co/ibm-granite/granite-4.0-h-micro-GGUF/tree/dc1dd2585fac18a78001c677d33ef8a7bbb7eb68))
```sh
llama cli -hf ibm-granite/granite-4.0-h-micro-GGUF:Q4_K_M
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf ibm-granite/granite-4.0-h-micro-GGUF:Q4_K_M
```
### LM Studio
1. Install LM Studio. Open the app once before using the lms command. [Download it (macOS / Windows / Linux)](https://lmstudio.ai/download)
2. Run Granite 4.0 H Micro 3B (Q4_K_M, needs ~1.94 GB, from [huggingface.co/ibm-granite](https://huggingface.co/ibm-granite/granite-4.0-h-micro-GGUF/tree/dc1dd2585fac18a78001c677d33ef8a7bbb7eb68))
```sh
lms get https://huggingface.co/ibm-granite/granite-4.0-h-micro-GGUF@Q4_K_M -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Other versions: 4bit (~1.8 GB, Mac only) `lms get https://huggingface.co/mlx-community/granite-4.0-h-micro-4bit -y` · 8bit (~3.4 GB, Mac only) `lms get https://huggingface.co/mlx-community/granite-4.0-h-micro-8bit -y`
### Ollama
1. Install Ollama. [Download it (macOS / Windows)](https://ollama.com/download)
```sh
# Linux
curl -fsSL https://ollama.com/install.sh | sh
```
2. Run Granite 4.0 H Micro 3B (Q4_K_M, needs ~1.9 GB, from [ollama.com](https://ollama.com/library/granite4/tags))
```sh
ollama run granite4:micro-h
```
### MLX (Mac)
1. Install MLX (Mac). Apple silicon Macs only, text only.
```sh
# with uv
uv tool install mlx-lm
# or
pip install mlx-lm
```
2. Run Granite 4.0 H Micro 3B (4bit, needs ~1.8 GB, Mac only, from [huggingface.co/mlx-community](https://huggingface.co/mlx-community/granite-4.0-h-micro-4bit/tree/0a29e17503da7de371af61a0a532853810637627))
```sh
mlx_lm.chat --model mlx-community/granite-4.0-h-micro-4bit
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model mlx-community/granite-4.0-h-micro-4bit
```
Other versions: 8bit (~3.4 GB, Mac only) `mlx_lm.chat --model mlx-community/granite-4.0-h-micro-8bit`
Model repository and documentation: [huggingface.co/ibm-granite](https://huggingface.co/ibm-granite/granite-4.0-h-micro)
[interactive version](https://whatmodel.sh/model/granite-4-0-h-micro) · [JSON](https://whatmodel.sh/api?format=json&all=1)