whatmodel.sh / Granite 4.0 H 1B

Granite 4.0 H 1B

+ contribute

Granite 4.0 H 1B is an open-weights language model from IBM, released October 2025. Its published size is 1.46B parameters and a 131k-token context window. Licence: apache-2.0.

Catalogue identity notes

IBM granite-4.0-h-1b (ibm-granite/granite-4.0-h-1b @d18cca4, created 2025-10-07, apache-2.0, 1,461,538,368 params): hybrid Mamba-2 dense instruct model of the Granite 4.0 Nano family (Oct 2025); distinct from the transformer-only granite-4.0-1b already catalogued. Lab rows from the h-1b card's 'H 1B Dense' column, in 4-model runs with the catalogued granite-4-0-1b / granite-4-0-350m columns.

Model information sources

Release date ↗ · Weights licence ↗

HF safetensors metadata of ibm-granite/granite-4.0-h-1b @d18cca4

40.4±0.8#527 of 719 LLMsfrom 29 benchmark results across 2 sources

$ how it compares

The LLMs ranked either side of Granite 4.0 H 1B, out of 828 with a score, then more from IBM.

#	model	score	where	RAM	released
617	Molmo 7B-D (0924) Ai2	41±1	local	~35GB	2024-09
618	OpenHermes-2.5-Code-290k-13B ajibawa-2023	41±1	local	~9.8GB	2024-03
619	Granite 4.0 H 1B IBM	40±1	local	~2.3GB	2025-10
620	Llama3-OpenBioLLM-8B Saama AI Labs	40±2	local	~6.6GB	2024-04
621	Nous Hermes 2 Mistral 7B DPO Nous Research	40±2	local	~6.1GB	2024-02
	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▁▁▃▆█▆▅▅▅▄▃▁ #245 of 300 * · 1.8% of its score · 1 test

Beats DeepSeek R1 0528 Qwen3 8B, Magistral Small 1.0 (2506), Llama 3.3 70B Instruct here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
BFCL v3 +1	50.7% ✎	▄▄·······▄██ beats 3/6	medium · avg 61%	▇ 1.8%	blunt · at its level · several runs

coding█··········█ #1 of 4 · 26% of its score · 7 tests

Beats Granite 4.0 1B here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
HumanEval lab	73.0% ✎	▁▁▁▁▁▃▅▅█▃▂▁ beats 133/190	very easy · avg 90%	▇▇▇ 6.8%	sharp · at its level
HumanEval+ lab	68.0% ✎	▁·▁▁▂▃▄▄█▅▄▃ beats 73/145	very easy · avg 88%	▇▇ 6.2%	sharp · at its level
BigCodeBench lab	29.1% ✎	▅····▃·█▃▅█▃ beats 4/12	medium · avg 42%	▇ 3.2%	at its level
MultiPL-E lab	36.1% ✎	▃··▄·▅▅▇▄█▃▄ beats 4/31	medium · avg 54%	▇ 3.1%	at its level
MBPP lab	69.0% ✎	▁▁▁▄▅█▄▅▃▁▂▁ beats 62/81	easy · avg 82%	▇ 2.8%	at its level
MBPP+ lab	60.0% ✎	▁··▁▁▃▄█▆▄▁▁ beats 42/120	easy · avg 73%	▇ 2.2%	at its level
CRUXEval-O lab	36.0% ✎	▂▂██▅▂▁·▁·▁▁ beats 17/51	medium · avg 51%	▇ 1.6%	blunt · at its level

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #237 of 393 · 6.2% of its score · 3 tests

Beats Mistral Small 4 (119B-A6.5B), Ministral 3 14B Reasoning 2512, Qwen2.5 Coder 32B Instruct here, though they rank higher overall. Trails Flan-T5 XXL here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
IFEval mean of four measures	80.1% ✎	█·█··█·██·██ beats 4/6	very easy · avg 87%	▇ 2.6%	typical
IFEval prompt-level strict lab	74.7% ✎	▁▁··▂··▃▇█▃▃ beats 5/26	easy · avg 81%	▇ 1.9%	at its level
IFEval instruction-level strict lab	82.4% ✎	▄▄····▄▄█▄·█ beats 3/8	very easy · avg 89%	▇ 1.8%	typical

math▁▂▃▅▆▅▄█▄▁▁▁ #325 of 398 · 29% of its score · 7 tests

Beats Athene-70B, GPT-3.5 Turbo (0125), Qwen2-VL 7B Instruct here, though they rank higher overall. Trails Gemma 7B (base), DeepSeek R1 Distill Qwen 1.5B, Tulu 2 DPO 13B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
GSM8K +1	69.6% ✎	█····▄▄▄···█ beats 3/6	very easy · avg 87%	▇▇▇▇▇ 12%	sharp · at its level · several runs
DeepMind Mathematics lab	35.0% ✎	▅····▅·█▃▃·▅ beats 3/10	medium · avg 54%	▇▇ 4.5%	at its level
GSM-Symbolic lab	65.7% ✎	▃▃····▃·▅▅█▃ beats 2/10	easy · avg 80%	▇ 3.0%	at its level
Minerva MATH lab	49.4% ✎	▄·▂▂·██▆·▆▄▄ beats 14/21	medium · avg 62%	▇ 2.9%	at its level
MGSM, 5 languages (en, es, fr, ja, zh) lab	27.5% ✎	█·▄▄██···█·▄ beats 2/10	medium · avg 49%	▇ 2.2%	at its level
MATH-500	47.2% ✎	▄··█▄·▄···▄▄ beats 3/6	easy · avg 80%	▇ 2.1%	at its level
AIME 2025	1.0% ✎	▄█··▄···▄··█ beats 1/6	hard · avg 33%	▇ 1.9%	sharp · noisy

other tests8 not in a task · 37% of its score

test	result	vs the others in its run	difficulty	counts	why
BIG-Bench Hard lab	59.7% ✎	▁▁▁▁▂▄▅▅█▃▄▁ beats 65/129	easy · avg 78%	▇▇▇ 8.5%	sharp · at its level
GPQA (subset unspecified) lab	29.7% ✎	▂▁▂▂▄▃▆▃▅█▄▃ beats 19/104	hard · avg 38%	▇▇▇ 7.2%	sharp · too hard for it
MMLU-Pro +1	27.6% ✎	█···█·██·███ beats 2/6	medium · avg 49%	▇▇ 5.9%	several runs
MMLU lab	59.7% ✎	▁▁▁▁▁▂▃▅█▃▆▃ beats 68/183	easy · avg 75%	▇▇ 5.7%	at its level
INCLUDE lab	43.4% ✎	▂····▃▂▄▄▄█▄ beats 4/36	medium · avg 54%	▇ 3.2%	typical
MMMLU lab	49.4% ✎	▂▁▁▂▃▃▄▄▅██▃ beats 10/46	medium · avg 58%	▇ 2.4%	at its level
IFBench (aggregation unspecified)	24.9% ✎	▄·█·█······█ beats 3/6	medium · avg 43%	▇ 2.1%	at its level
GPQA Diamond	24.3% ✎	█··········█ beats 0/5	hard · avg 41%	▇ 2.0%	sharp · too hard for it · noisy

$ where to run it

llama.cpp · recommended

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 H 1B · needs ~0.9 GB · from huggingface.co/ibm-granite
chat in the terminal
llama cli -hf ibm-granite/granite-4.0-h-1b-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-h-1b-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 H 1B · needs ~0.9 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-h-1b-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 H 1B · needs ~1.6 GB · from ollama.com
chat in the terminal
ollama run granite4:1b-h

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 H 1B · needs ~0.8 GB, Mac only · from huggingface.co/mlx-community
chat in the terminal
mlx_lm.chat --model mlx-community/granite-4.0-h-1b-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model mlx-community/granite-4.0-h-1b-4bit

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

$ how the score was worked out · ibm-granite/granite-4.0-h-1b… 80%, LiquidAI/LFM2.5-1.2B-Thinking… 20%

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-1b model card · lab-reported	21	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 80%	Apache-2.0
LiquidAI/LFM2.5-1.2B-Thinking model card	8	▇▇▇▇▇▇ 20%	LFM Open License v1.0
prior · pull toward the middle		▇ 0%
weighted average		= 40.4

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated 2026-10-11 02:20 UTC · sources and licences · json