Granite 4.1 8B

+ contribute

Granite 4.1 8B is an open-weights language model from IBM, released April 2026. Its published size is 8.79B parameters and a 131k-token context window. Licence: apache-2.0.

Catalogue identity notes

Comparison-only. Model card (ibm-granite/granite-4.1-8b @1504002): 8B instruct model, Release Date April 29th 2026, Apache 2.0; parameter count 8,791,592,960 from the Hugging Face safetensors metadata (dense, so active = total). Arena entry 'granite-4.1-8b' (ibm, Apache 2.0).

54.3±0.9#282 of 719 LLMsfrom 24 benchmark results across 3 sources

$ how it compares

The LLMs ranked either side of Granite 4.1 8B, out of 828 with a score, then more from IBM.

#	model	score	where	RAM	released
334	Llama 4 Scout (109B-A17B) Meta	54±1	api $0.15/M +local	~36GB	2025-04
335	Nemotron 3 Nano 4B NVIDIA	54±1	local	~3.8GB	2026-03
336	Granite 4.1 8B IBM	54±1	api $0.11/M +local	~7GB	2026-04
337	Pixtral Large (2411) Mistral	54±3	local	~51GB	2024-11
338	Bielik 11B v2.6 Instruct SpeakLeash	54±3	local	~8.1GB	2025-06
	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▁▁▃▆█▆▅▅▅▄▃▁ #206 of 300 * · 7.6% of its score · 2 tests

Beats o1 (2024-12-17), o3-mini, DeepSeek R1 (Jan '25) here, though they rank higher overall. Trails SmolLM3 3B, Jamba 1.5 Mini, Mellum2 12B-A2.5B Instruct here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
τ²-bench Telecom lab	17.5% ✎	▁·▅█▆▅▂▃▃▂▃▁ beats 10/89	very hard · avg 12%	▇▇▇ 6.5%	sharp
BFCL v3 lab	68.3% ✎	▁▁▁▁▁▃▅█▆▅▄▁ beats 141/170	medium · avg 61%	▇ 1.0%	blunt · at its level · small run

coding▁▁▁▃▆██▇▆▄▃▁ #193 of 458 · 36% of its score · 8 tests

Beats Gemini 2.5 Flash, Qwen3.5 4B, Qwen3 30B-A3B Instruct 2507 here, though they rank higher overall. Trails QwQ-32B-Preview, DeepSeek-Coder-V2 Instruct (0724), DeepSeek-V2.5 here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Arena WebDev (Code Arena)	1190.3 ✎	▁▃▄▃▅█▄▇▆▄▁▁ beats 4/119	avg 981	▇▇▇▇▇ 13%	sharp
HumanEval lab	85.4% ✎	▁▁▁▁▁▃▅▅█▃▂▁ beats 174/190	easy · avg 80%	▇▇ 6.1%	sharp · small run
HumanEval+ lab	79.9% ✎	▁·▁▁▂▃▄▄█▅▄▃ beats 109/145	easy · avg 76%	▇▇ 5.4%	sharp · small run
MBPP lab	87.3% ✎	▁▁▁▄▅█▄▅▃▁▂▃ beats 80/81	easy · avg 76%	▇ 2.6%	small run
MultiPL-E lab	60.3% ✎	▃··▄·▅▅▇▄█▃▄ beats 13/31	medium · avg 56%	▇ 2.5%	at its level · small run
BigCodeBench lab	35.0% ✎	▅····▃·█▃▅█▃ beats 7/12	hard · avg 32%	▇ 2.5%	at its level · small run
MBPP+ lab	73.8% ✎	▁··▁▁▃▄█▆▄▁▁ beats 96/120	easy · avg 67%	▇ 2.1%	at its level · small run
CRUXEval-O lab	47.6% ✎	▂▂██▅▂▁·▁·▁▁ beats 41/51	medium · avg 48%	▇ 1.4%	blunt · at its level · small run

chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #215 of 393 * · 8.9% of its score · 2 tests

Beats Granite 4.2 8B, Mistral Small 4 (119B-A6.5B), Gemma 2 9B IT SimPO here, though they rank higher overall. Trails LFM2.5 1.2B Instruct, Granite 4.0 H Micro 3B, Gemma 3n E4B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
LMArena (text)	1304.1 ✎	▁▁▁▂▃▃▃▆▆▇█▃ beats 131/379	avg 1268	▇▇▇ 7.2%	blunt
IFEval (aggregation unspecified) lab	87.1% ✎	▁▁▃▁▂▄▄▆█▃▂▁ beats 87/141	easy · avg 82%	▇ 1.7%	small run

math▄··········█ #1 of 3 · 17% of its score · 5 tests

Beats Granite 4.1 30B here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
GSM8K lab	92.5% ✎	▂▂▂▅▅▆▅▇█▇▄▃ beats 130/146	very easy · avg 88%	▇▇ 5.6%	sharp · maxed out · small run
Minerva MATH lab	80.1% ✎	▄·▂▂·██▆·▆▄▄ beats 20/21	easy · avg 68%	▇▇ 3.9%	small run
DeepMind Mathematics lab	80.1% ✎	▅····▅·█▃▃·▅ beats 9/10	easy · avg 67%	▇▇ 3.8%	small run
GSM-Symbolic lab	83.7% ✎	▃▃····▃·▅▅█▃ beats 9/10	easy · avg 78%	▇ 2.0%	small run
MGSM, 5 languages (en, es, fr, ja, zh) lab	82.3% ✎	█·▄▄██···█·▄ beats 10/10	medium · avg 63%	▇ 1.6%	at its level · small run

other tests7 not in a task · 31% of its score

test	result	vs the others in its run	difficulty	counts	why
GPQA (subset unspecified) lab	42.0% ✎	▂▁▂▂▄▃▆▃▅█▄▃ beats 43/104	hard · avg 37%	▇▇ 6.2%	sharp · small run
BIG-Bench Hard lab	80.5% ✎	▁▁▁▁▂▄▅▅█▃▄▁ beats 110/129	easy · avg 74%	▇▇ 6.1%	sharp · small run
MMLU lab	73.8% ✎	▁▁▁▁▁▂▃▅█▃▆▃ beats 124/183	medium · avg 70%	▇▇ 4.4%	at its level · small run
SimpleQA lab	4.8% ✎	▄▄▄▆█▆███·▂█ beats 4/32	very hard · avg 4.8%	▇▇ 3.9%	too hard for it · small run
MMLU-Pro lab	56.0% ✎	▁▁·▁▁▁▃▃▄▅█▁ beats 55/188	medium · avg 52%	▇ 3.7%	at its level · small run
INCLUDE lab	58.9% ✎	▂····▂▂▄▄▄█▄ beats 11/36	medium · avg 55%	▇ 3.5%	at its level · small run
MMMLU lab	64.8% ✎	▂▁▁▂▃▂▄▄▅██▃ beats 17/46	medium · avg 61%	▇ 2.9%	at its level · small run

$ where to run it

Through an API

api · Replicate · $0.11/M · $0.06 in / $0.25 out per 1M tokens · 131k context · replicate.com ↗ · price sourceUSD per 1M tokens, Replicate official model ibm-granite/granite-4.1-8b, billed per input/output token (billing table on the model page: $0.06 per million input tokens, $0.25 per million output tokens; A100 80GB backing hardware not billed separately for this official model). IBM watsonx lists granite-4-1-8b as not available for pay-per-token. Currency: USD. Replicate official model page, observed 2026-09-25.

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 4.1 8B · needs ~5.35 GB · from huggingface.co/ibm-granite
chat in the terminal
llama cli -hf ibm-granite/granite-4.1-8b-GGUF:Q4_K_M
Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf ibm-granite/granite-4.1-8b-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 4.1 8B · needs ~5.35 GB · from huggingface.co/ibm-granite
open in LM Studio ↗
Or download it in the terminal
lms get https://huggingface.co/ibm-granite/granite-4.1-8b-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 4.1 8B · needs ~5.35 GB · from ollama.com
chat in the terminal
ollama run granite4.1:8b

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 4.1 8B · needs ~5.2 GB, Mac only · from huggingface.co/mlx-community
chat in the terminal
mlx_lm.chat --model mlx-community/granite-4.1-8b-4bit
Or start an OpenAI-compatible server at http://localhost:8080/v1
mlx_lm.server --model mlx-community/granite-4.1-8b-4bit

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

$ how the score was worked out · ibm-granite/granite-4.1-8b mo… 73%, LMArena leaderboard dataset 20%, Synthesize and Reward: Reinfo… 7%

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-4.1-8b model card · lab-reported	21	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 73%	Apache-2.0
LMArena leaderboard dataset	2	▇▇▇▇▇▇ 20%	CC BY 4.0
Synthesize and Reward: Reinforcement Learning for Multi-Step Tool Use in Live Environments (arXiv:2606.03892v2) · lab-reported	1	▇▇ 7%	CC-BY-4.0
prior · pull toward the middle		▇ 0%
weighted average		= 54.3

$ 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