whatmodel.sh / Granite 4.0 Micro 3B

Granite 4.0 Micro 3B

+ contribute

Granite 4.0 Micro 3B is an open-weights language model from IBM, released September 2025. Its published size is 3.4B parameters and a 131k-token context window. Licence: apache-2.0.

46.0±0.9#420 of 719 LLMsfrom 27 benchmark results across 4 sources

$ how it compares

The LLMs ranked either side of Granite 4.0 Micro 3B, out of 828 with a score, then more from IBM.

#	model	score	where	RAM	released
498	Hermes 2 Pro Llama 3 8B Nous Research	46±3	local	~6.6GB	2024-04
499	Llama 3 70B (base) Meta	46±2	local	~43GB	2024-04
500	Granite 4.0 Micro 3B IBM	46±1	local	~3.5GB	2025-09
501	Command R (08-2024) Cohere	46±1	api $0.26/M +local	~21GB	2024-08
502	GPT-3.5 Turbo (0301) OpenAI	46±1	comparison only		2023-03
	more from IBM
192	Granite 4.2 30B IBM	63±1	api $0.28/M +local	~19GB	2026-08
239	Granite 4.2 8B IBM	60±1	api $0.11/M +local	~6.6GB	2026-08
276	Granite 4.2 3B IBM	57±1	api $0.052/M +local	~3.7GB	2026-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▄·▄········█ #2 of 4 * · 1.2% of its score · 1 test

Beats Granite 4.0 H Tiny (7B-A1B), Granite 4.0 H Micro 3B here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
BFCL v3 lab	60.0% ✎	▁▁▁▁▁▃▅█▆▅▄▁ beats 99/170	medium · avg 60%	▇ 1.2%	blunt · at its level

coding█···▄······▄ #3 of 4 · 32% of its score · 7 tests

Beats Granite 4.0 H Micro 3B here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
HumanEval lab	80.0% ✎	▁▁▁▁▁▃▅▅█▃▂▁ beats 161/190	very easy · avg 86%	▇▇▇▇ 10%	sharp
HumanEval+ lab	72.0% ✎	▁·▁▁▂▃▄▄█▅▄▃ beats 90/145	easy · avg 81%	▇▇▇ 8.4%	sharp · at its level
MBPP lab	72.0% ✎	▁▁▁▄▅█▄▅▃▁▂▁ beats 68/81	easy · avg 79%	▇▇ 3.8%	at its level
MultiPL-E lab	49.2% ✎	▃··▄·▅▅▇▄█▃▄ beats 5/31	medium · avg 58%	▇ 2.9%	at its level
BigCodeBench lab	39.2% ✎	▅····▃·█▃▅█▃ beats 10/12	medium · avg 42%	▇ 2.9%	at its level
MBPP+ lab	64.0% ✎	▁··▁▁▃▄█▆▄▁▁ beats 60/120	easy · avg 70%	▇ 2.6%	at its level
CRUXEval-O lab	41.5% ✎	▂▂██▅▂▁·▁·▁▁ beats 33/51	medium · avg 48%	▇ 1.6%	blunt · at its level

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #222 of 393 * · 4.6% of its score · 2 tests

Beats Granite 4.2 8B, Mistral Small 4 (119B-A6.5B), Olmo 3.1 32B Think here, though they rank higher overall. Trails LFM2.5 1.2B Instruct, Gemma 3 4B, Olmo 3 7B Instruct here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
IFEval instruction-level strict lab	85.5% ✎	▄▄····▄▄█▄·█ beats 5/8	very easy · avg 88%	▇ 2.4%	typical
IFEval prompt-level strict lab	79.1% ✎	▁▁··▂··▂▇█▃▃ beats 8/26	easy · avg 81%	▇ 2.2%	typical

math▁▂▃▅▆▅▄█▄▁▁▁ #272 of 398 · 24% of its score · 6 tests

Beats Molmo 2 8B, DeepSeek-Coder-V2 Instruct (0724), Llama 3.1 70B Instruct here, though they rank higher overall. Trails DeepSeek R1 Distill Qwen 1.5B, Mixtral 8x7B v0.1 (base), InternVL2-4B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
GSM8K lab	85.5% ✎	▂▂▂▅▅▆▅▇█▇▄▃ beats 105/146	very easy · avg 87%	▇▇▇ 7.0%	sharp
DeepMind Mathematics lab	44.6% ✎	▅····▅·█▃▃·▅ beats 5/10	medium · avg 55%	▇▇ 5.0%	at its level
Minerva MATH lab	62.1% ✎	▄·▂▂·██▆·▆▄▄ beats 15/21	easy · avg 70%	▇▇ 4.9%	at its level
GSM-Symbolic lab	79.8% ✎	▃▃····▃·▅▅█▃ beats 6/10	easy · avg 83%	▇▇ 4.3%	typical
MGSM, 5 languages (en, es, fr, ja, zh) lab	28.6% ✎	█·▄▄██···█·▄ beats 3/10	medium · avg 46%	▇ 1.6%	at its level
OTIS mock AIME 2024–2025	2.8% ✎	▃▃▄▃▃▃▇▇█▃▂▅ beats 23/193	very hard · avg 8.5%	▇ 1.0%	sharp · too hard for it · noisy

other tests11 not in a task · 38% of its score

test	result	vs the others in its run	difficulty	counts	why
BIG-Bench Hard lab	72.5% ✎	▁▁▁▁▂▄▅▅█▃▄▁ beats 101/129	easy · avg 76%	▇▇▇ 7.0%	sharp · at its level
GPQA (subset unspecified) lab	30.1% ✎	▂▁▂▂▄▃▆▃▅█▄▃ beats 22/104	hard · avg 37%	▇▇ 5.8%	sharp
MMLU lab	66.0% ✎	▁▁▁▁▁▂▃▅█▃▆▃ beats 92/183	easy · avg 74%	▇▇ 5.6%	at its level
GPQA Diamond	28.3% ✎	▁▁▁▃▂▄▆▄▅▇█▅ beats 14/208	hard · avg 40%	▇▇ 5.5%	sharp
MMLU-Pro lab	44.5% ✎	▁▁·▁▁▁▃▂▄▅█▁ beats 38/188	medium · avg 52%	▇▇ 4.3%	at its level
INCLUDE lab	51.6% ✎	▂····▂▃▄▄▄█▄ beats 6/36	medium · avg 58%	▇▇ 3.9%	at its level
MMMLU lab	55.1% ✎	▂▁▁▂▃▂▄▄▅██▃ beats 11/46	medium · avg 63%	▇ 3.4%	at its level
EuroEval HellaSwag (English), test split, 5-shot generative MC, accuracy	70.8% ✎	▃·▁▁▂▁▂▅█▇▄▁ beats 52/100	medium · avg 68%	▇ 1.2%	at its level · noisy

3 more tests

test	result	vs the others in its run	difficulty	counts	why
EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy	77.3% ✎	▁▁▁▁▁▁▁▂▄█▄▃ beats 57/146	easy · avg 83%	▇ 0.9%	at its level · noisy
Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate)	93.5% ✎	▁▁▂▃▄▄█▆▃▂▂▁ beats 47/140	very easy · avg 94%	▇ 0.5%	blunt · maxed out · noisy
Chess Puzzles (Epoch)	0.0% ✎	█▁▁▁▁▂▁▄▂▂▁▁ beats 45/135	very hard · avg 5.9%	▇ 0.1%	sharp · too hard for it · noisy

$ where to run it

llama.cpp · recommended

version

Install llama.cpp · Homebrew and winget installs name the commands llama-cli and llama-server instead.
macOS / Linux
curl -LsSf https://llama.app/install.sh | sh
Windows
irm https://llama.app/install.ps1 | iex
other ways to install ↗
Run Granite 4.0 Micro 3B · needs ~2.1 GB · from huggingface.co/ibm-granite
chat in the terminal
llama cli -hf ibm-granite/granite-4.0-micro-GGUF:Q4_K_M
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf ibm-granite/granite-4.0-micro-GGUF:Q4_K_M

LM Studio

version

Install LM Studio · Open the app once before using the lms command.
download it (macOS / Windows / Linux) ↗
Run Granite 4.0 Micro 3B · needs ~2.1 GB · from huggingface.co/ibm-granite
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/ibm-granite/granite-4.0-micro-GGUF@Q4_K_M -y
Then run lms ls to see its name, and lms chat followed by that name

Ollama

Install Ollama
download it (macOS / Windows) ↗
Linux
curl -fsSL https://ollama.com/install.sh | sh
Run Granite 4.0 Micro 3B · needs ~2.54 GB · from ollama.com
chat in the terminal
ollama run granite4:micro

MLX (Mac)

version

Install MLX (Mac) · Apple silicon Macs only, text only.
with uv
uv tool install mlx-lm
or
pip install mlx-lm
Run Granite 4.0 Micro 3B · needs ~1.9 GB, Mac only · from huggingface.co/ethicalabs
chat in the terminal
mlx_lm.chat --model ethicalabs/granite-4.0-micro-MLX
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model ethicalabs/granite-4.0-micro-MLX

Weights and documentation: huggingface.co/ibm-granite ↗

$ how the score was worked out · ibm-granite/granite-4.0-h-sma… 90%, Epoch AI — internal evaluatio… 7%, EuroEval leaderboards (result… 2% and 1 more

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
ibm-granite/granite-4.0-h-small model card · lab-reported	21	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 90%	Apache-2.0
Epoch AI — internal evaluations	3	▇▇ 7%	CC BY 4.0
EuroEval leaderboards (results collection), English generative leaderboard	2	▇ 2%	MIT
Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository)	1	▇ 0%	Apache-2.0
prior · pull toward the middle		▇ 0%
weighted average		= 46.0