[whatmodel.sh](../../README.md) / Granite 4.1 8B
# Granite 4.1 8B
IBM · released 2026-04 · 8.79B params · apache-2.0
**score 55.2** ±0.9 · #276 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 4.1 8B among every model on that task or test, worst to best.
task vs the others rank counts tests
agent tasks ▁▁▃▆●▄▅▅▄▄▂▁ #219 of 313* 7.9% 2
coding ▁▁▁▃▅▇█●▅▄▃▁ #203 of 469 35% 8
chat & writing ▁▁▁▃▆▄▆●▆▇▇▃ #223 of 402* 9.3% 2
math ▄··········● #1 of 3 17% 5
agent tasks · #219 of 313 · 7.9% of its score
Beats [o1 (2024-12-17)](o1-2024-12-17.md), [o3-mini](o3-mini.md), [DeepSeek R1 (Jan '25)](deepseek-r1.md) here, though they rank higher overall. Trails [SmolLM3 3B](smollm3-3b.md), [Jamba 1.5 Mini](jamba-1-5-mini.md), [Mellum2 12B-A2.5B Instruct](mellum2-12b-a2-5b-instruct.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
τ²-bench Telecom (maker) 17.5% ▁·●█▆▅▂▃▃▂▃▁ beats 10/89 very hard · avg 11% 6.9% sharp
BFCL v3 (maker) 68.3% ▁▁▁▁▁▃▅█▆●▄▁ beats 142/171 medium · avg 60% 1.0% blunt · at its level · small run
coding · #203 of 469 · 35% of its score
Beats [Gemini 2.5 Flash](gemini-2-5-flash.md), [Qwen3 30B-A3B Instruct 2507](qwen3-30b-a3b-instruct-2507.md), [Qwen3.5 4B](qwen3-5-4b.md) here, though they rank higher overall. Trails [QwQ-32B-Preview](qwq-32b-preview.md), [Ministral 3 3B Reasoning 2512](ministral-3-3b-reasoning.md), [DeepSeek-Coder-V2 Instruct (0724)](deepseek-coder-v2-0724.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
Arena WebDev (Code Arena) 1192.5 ▁●▃▂▅█▃▆▅▅▁▂ beats 4/129 avg 981 13% sharp
HumanEval (maker) 85.4% ▁▁▁▁▂▂▆▆█●▂▁ beats 189/205 easy · avg 77% 5.8% sharp · small run
HumanEval+ (maker) 79.9% ▁·▁▁▂▃▄▄█●▄▃ beats 109/145 easy · avg 73% 5.2% sharp · small run
MBPP (maker) 87.3% ▁▁▁▄▅█▆▅▄▁▂● beats 85/86 easy · avg 74% 2.5% small run
MultiPL-E (maker) 60.3% ▃··▄·▅▅●▄█▃▄ beats 13/31 medium · avg 54% 2.4% at its level · small run
BigCodeBench (maker) 35.0% ▅····▃·█▃●█▃ beats 7/12 hard · avg 30% 2.3% at its level · small run
MBPP+ (maker) 73.8% ▁··▁▁▃▄█▆●▁▁ beats 96/120 medium · avg 65% 2.0% at its level · small run
CRUXEval-O (maker) 47.6% ▂▂██●▂▁·▁·▁▁ beats 41/51 medium · avg 46% 1.3% blunt · at its level · small run
chat & writing · #223 of 402 · 9.3% 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), [Granite 4.0 H Micro 3B](granite-4-0-h-micro.md), [Gemma 3n E4B](gemma-3n-e4b.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
LMArena (text) 1304.0 ▁▁▁▂▃▃▃●▆▇█▃ beats 131/385 avg 1260 7.6% blunt
IFEval (aggregation unspecified) … 87.1% ▁▂▄▂▂▄▆●█▅▂▁ beats 87/145 easy · avg 81% 1.7% small run
math · #1 of 3 · 17% of its score
Beats [Granite 4.1 30B](granite-4-1-30b.md) here, though they rank higher overall.
test result vs the others in its run difficulty counts why
GSM8K (maker) 92.5% ▂▂▂▂▃▄▅▇██●▄ beats 185/206 very easy · avg 85% 6.7% sharp · maxed out · small run
Minerva MATH (maker) 80.1% ▁·▃▁▁▂▁█▄▁▃● beats 26/27 easy · avg 66% 3.9% small run
DeepMind Mathematics (maker) 80.1% ▅····▅·█▃▃·● beats 9/10 easy · avg 66% 3.5% small run
GSM-Symbolic (maker) 83.7% ▃▃····▃·▅▅●▃ beats 9/10 easy · avg 76% 1.9% small run
MGSM, 5 languages (en, es, fr, ja… 82.3% █·▄▄██···█·● beats 10/10 medium · avg 62% 1.5% at its level · small run
other tests (7)
test result vs the others in its run difficulty counts why
GPQA (subset unspecified) (maker) 42.0% ▂▁▂▂▄▃●▃▅█▃▃ beats 43/106 very hard · avg 35% 6.1% sharp · small run
BIG-Bench Hard (maker) 80.5% ▁▁▁▁▂▄▅▆█●▃▁ beats 136/155 medium · avg 72% 5.8% sharp · small run
MMLU (maker) 73.8% ▁▁▁▁▁▂▂▅█●▆▄ beats 152/214 medium · avg 67% 4.8% at its level · small run
MMLU-Pro (maker) 56.0% ▁▁·▁▁▁▃●▄▅█▁ beats 55/188 medium · avg 49% 3.7% at its level · small run
SimpleQA (maker) 4.8% ▄▄●▆█▆███·▂█ beats 4/32 very hard · avg 4.2% 3.7% too hard for it · small run
INCLUDE (maker) 58.9% ▂····▂▂●▄▄█▄ beats 11/36 medium · avg 54% 3.2% at its level · small run
MMMLU (maker) 64.8% ▂▁▁▂▃▂▄●▅██▃ beats 17/46 medium · avg 60% 2.9% at its level · small run
source results share of the score
ibm-granite/granite-4.1-8b model card (… 21 72% ▇▇▇▇▇▇▇▇▇
LMArena leaderboard dataset 2 21% ▇▇▇
Synthesize and Reward: Reinforcement Le… 1 7% ▇
prior: pull toward the middle 0% ▇
weighted average = 55.2
Licences: ibm-granite/granite-4.1-8b model card ([Apache-2.0](../../licenses/ibm-granite--granite-4.1-8b.txt)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · Synthesize and Reward: Reinforcement Learning for Multi-Step Tool Use in Live Environments ([CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/))
## How to run it
### Through an API
api (Replicate) · $0.11/M blended · $0.06 in / $0.25 out per 1M tokens · [replicate.com](https://replicate.com/ibm-granite/granite-4.1-8b)
[price source](https://replicate.com/ibm-granite/granite-4.1-8b): USD 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. · [context source](https://huggingface.co/ibm-granite/granite-4.1-8b/blob/1504002f650e656a0a3789d99574df12e3e94ed0/README.md): IBM official model card (ibm-granite/granite-4.1-8b, pinned): architecture table 'Sequence length 131072' for the 8B Dense model; the card calls it a long-context instruct model. Observed 2026-09-25.
### 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 4.1 8B (Q4_K_M, needs ~5.35 GB, from [huggingface.co/ibm-granite](https://huggingface.co/ibm-granite/granite-4.1-8b-GGUF/tree/865b82c2e7970d82e3731278c88c57ae7138359c))
```sh
llama cli -hf ibm-granite/granite-4.1-8b-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-4.1-8b-GGUF:Q4_K_M
```
Other versions: Q8_0 (~9.35 GB) `llama cli -hf ibm-granite/granite-4.1-8b-GGUF:Q8_0`
### 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 4.1 8B (Q4_K_M, needs ~5.35 GB, from [huggingface.co/ibm-granite](https://huggingface.co/ibm-granite/granite-4.1-8b-GGUF/tree/865b82c2e7970d82e3731278c88c57ae7138359c))
```sh
lms get https://huggingface.co/ibm-granite/granite-4.1-8b-GGUF@Q4_K_M -y
```
3. Then run lms ls to see its name, and lms chat followed by that name
Other versions: Q8_0 (~9.35 GB) `lms get https://huggingface.co/ibm-granite/granite-4.1-8b-GGUF@Q8_0 -y` · 4bit (~5.2 GB, Mac only) `lms get https://huggingface.co/mlx-community/granite-4.1-8b-4bit -y` · 8bit (~9.4 GB, Mac only) `lms get https://huggingface.co/mlx-community/granite-4.1-8b-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 4.1 8B (Q4_K_M, needs ~5.35 GB, from [ollama.com](https://ollama.com/library/granite4.1/tags))
```sh
ollama run granite4.1:8b
```
### 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 4.1 8B (4bit, needs ~5.2 GB, Mac only, from [huggingface.co/mlx-community](https://huggingface.co/mlx-community/granite-4.1-8b-4bit/tree/08fb1e272f7bd49fa83ce279bbdc496c980380ac))
```sh
mlx_lm.chat --model mlx-community/granite-4.1-8b-4bit
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model mlx-community/granite-4.1-8b-4bit
```
Other versions: 8bit (~9.4 GB, Mac only) `mlx_lm.chat --model mlx-community/granite-4.1-8b-8bit`
Model repository and documentation: [huggingface.co/ibm-granite](https://huggingface.co/ibm-granite/granite-4.1-8b)
## Published speed measurements
Reference results for the hardware, engine and settings below. These are not verified speeds of the downloads above or predictions for your machine.
hardware precision speed engine and settings
Apple M5 Max (40-core GPU),… q4_k_m 69.65 tok/s ollama (version not recorded) · l…
Each speed links to its source, which has the full settings.
[interactive version](https://whatmodel.sh/model/granite-4-1-8b) · [JSON](https://whatmodel.sh/api?format=json&all=1)