whatmodel.sh / Granite 4.0 H 350M

Granite 4.0 H 350M

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Granite 4.0 H 350M is an open-weights language model from IBM, released October 2025. Its published size is 0.34B parameters and a 33k-token context window. Licence: apache-2.0.

Catalogue identity notes

IBM granite-4.0-h-350m (ibm-granite/granite-4.0-h-350m @3b17b71, created 2025-10-07, apache-2.0, 340,332,224 params): hybrid Mamba-2 dense instruct model of the Granite 4.0 Nano family (Oct 2025); distinct from the transformer-only granite-4.0-350m already catalogued. Lab rows from the h-1b card's 'H 350M 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-350m @3b17b71

26.8±1.1#662 of 719 LLMsfrom 21 benchmark results across 1 source

$ how it compares

The LLMs ranked either side of Granite 4.0 H 350M, out of 828 with a score, then more from IBM.

#modelscorewhereRAMreleased
761RedPajama-INCITE 7B Base Together27±1local~16GB2023-06
762ChatGLM3-6B Zhipu AI (THUDM)27±2local~5.8GB2023-10
763Granite 4.0 H 350M IBM27±1local~1.7GB2025-10
764Flan-T5 Large Google27±2local~4.8GB2022-10
765MPT-7B-Chat MosaicML27±2comparison only2023-05
more from IBM
192Granite 4.2 30B IBM63±1api $0.28/M +local~19GB2026-08
239Granite 4.2 8B IBM60±1api $0.11/M +local~6.6GB2026-08
276Granite 4.2 3B IBM57±1api $0.052/M +local~3.7GB2026-08

see all LLMs · find the ones that fit your machine

$ what it's good at

agent tasks▁▁▃▆█▆▅▅▅▄▃▁ #268 of 300 * · 1.7% of its score · 1 test

Beats Magistral Small 1.0 (2506), Llama 4 Scout (109B-A17B), Bielik 11B v2.3 Instruct here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
BFCL v3 lab43.3% ✎▁▁▁▁▁▃▅█▆▅▄▁ beats 23/170medium · avg 61%▇ 1.7%blunt · at its level
coding█··········█ #4 of 4 · 40% of its score · 7 tests

Trails Granite 4.0 350M here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
HumanEval lab38.0% ✎▁▁▁▁▁▃▅▅█▃▂▁ beats 50/190very easy · avg 90%▇▇▇▇▇ 12%sharp · at its level
HumanEval+ lab35.0% ✎▁·▁▁▂▃▄▄█▅▄▃ beats 24/145very easy · avg 88%▇▇▇▇ 10%sharp · at its level
MBPP lab49.0% ✎▁▁▁▄▅█▄▅▃▁▂▁ beats 28/81easy · avg 82%▇▇ 5.0%at its level
MultiPL-E lab14.3% ✎▃··▄·▅▅▇▄█▃▄ beats 0/31medium · avg 54%▇ 3.5%typical
MBPP+ lab44.0% ✎▁··▁▁▃▄█▆▄▁▁ beats 12/120easy · avg 73%▇ 3.4%at its level
BigCodeBench lab11.2% ✎▅····▃·█▃▅█▃ beats 1/12medium · avg 42%▇ 3.3%typical
CRUXEval-O lab25.5% ✎▃▂██▅▂▁·▁·▁▁ beats 2/51medium · avg 51%▇ 2.0%blunt
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #330 of 393 * · 7.4% of its score · 2 tests

Beats OpenThinker3 7B, Gemma 4 E2B, Code Llama 70B Instruct here, though they rank higher overall.

testresultvs the others in its rundifficultycountswhy
IFEval instruction-level strict lab67.6% ✎▄▄····▄▄█▄·█ beats 1/8very easy · avg 89%▇▇ 3.9%at its level
IFEval prompt-level strict lab55.6% ✎▁▁··▃··▂▇█▃▃ beats 3/26easy · avg 81%▇ 3.5%at its level
math▁▃▃▅▆▅▄█▄▁▁▁ #392 of 398 · 25% of its score · 5 tests

Beats Llama 2 13B Chat, DeepSeek-Coder 1.3B Base, StarCoderBase 7B here, though they rank higher overall. Trails Qwen2 0.5B Instruct here, though they rank lower overall.

testresultvs the others in its rundifficultycountswhy
GSM8K lab39.3% ✎▂▂▂▅▅▆▅▇█▇▄▃ beats 32/146very easy · avg 89%▇▇▇▇ 11%sharp · at its level
GSM-Symbolic lab33.7% ✎▃▃····▃·▅▅█▃ beats 1/10easy · avg 80%▇▇ 4.9%at its level
DeepMind Mathematics lab6.2% ✎▅····▅·█▃▃·▅ beats 0/10medium · avg 54%▇▇ 4.2%typical
Minerva MATH lab5.8% ✎▄·▂▂·██▆·▆▄▄ beats 1/21medium · avg 62%▇ 3.0%typical
MGSM, 5 languages (en, es, fr, ja, zh) lab16.2% ✎█·▄▄██···█·▄ beats 1/10medium · avg 49%▇ 2.6%typical
other tests6 not in a task · 25% of its score
testresultvs the others in its rundifficultycountswhy
BIG-Bench Hard lab33.3% ✎▁▁▁▁▃▄▅▅█▃▄▁ beats 12/129easy · avg 78%▇▇▇ 7.8%sharp
MMLU lab36.2% ✎▁▁▁▁▁▃▃▅█▃▆▃ beats 14/183easy · avg 75%▇▇ 6.2%typical
GPQA (subset unspecified) lab26.1% ✎▂▃▂▂▄▃▆▃▅█▄▃ beats 7/104hard · avg 38%▇ 3.6%sharp · too hard for it
MMLU-Pro lab14.4% ✎▁▃·▁▁▁▃▂▄▅█▁ beats 2/188medium · avg 49%▇ 2.9%typical
INCLUDE lab27.1% ✎▃····▂▂▄▄▄█▄ beats 0/36medium · avg 54%▇ 2.6%too hard for it
MMMLU lab28.0% ✎▃▁▁▂▃▂▄▄▅██▃ beats 0/46medium · avg 58%▇ 2.3%typical

$ where to run it

llama.cpp · recommended

  1. 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 ↗

  2. Run Granite 4.0 H 350M · needs ~0.22 GB · from huggingface.co/ibm-granite
    chat in the terminal
    llama cli -hf ibm-granite/granite-4.0-h-350m-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf ibm-granite/granite-4.0-h-350m-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run Granite 4.0 H 350M · needs ~0.22 GB · from huggingface.co/ibm-granite

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/ibm-granite/granite-4.0-h-350m-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run Granite 4.0 H 350M · needs ~0.366 GB · from ollama.com
    chat in the terminal
    ollama run granite4:350m-h

MLX (Mac)

version

  1. Install MLX (Mac) · Apple silicon Macs only, text only.
    with uv
    uv tool install mlx-lm
    or
    pip install mlx-lm
  2. Run Granite 4.0 H 350M · needs ~0.2 GB, Mac only · from huggingface.co/mlx-community
    chat in the terminal
    mlx_lm.chat --model mlx-community/granite-4.0-h-350m-4bit
  3. Or start an OpenAI-compatible server at http://localhost:8080/v1
    mlx_lm.server --model mlx-community/granite-4.0-h-350m-4bit

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

$ how the score was worked out · ibm-granite/granite-4.0-h-1b… 99%

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

sourceresultsshare of the scorelicence
ibm-granite/granite-4.0-h-1b model card · lab-reported21▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 99%Apache-2.0
prior · pull toward the middle▇ 1%
weighted average= 26.8

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit 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 · sources and licences · json