[whatmodel.sh](../../README.md) / Granite 3.0 2B Instruct
# Granite 3.0 2B Instruct
IBM · released 2024-10 · 2.63B params · apache-2.0
**score 40.7** ±0.8 · #553 of 732 · from 22 benchmark results across 5 sources
Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only.
## What it's good at
● marks Granite 3.0 2B Instruct among every model on that task or test, worst to best.
task vs the others rank counts tests
coding ▁▁▁▃●▇█▇▅▄▃▁ #376 of 469 18% 6
chat & writing ▁▁▁▃●▄▆█▆▇▇▃ #348 of 402* 10% 2
math ▁▁▃●█▆▄█▄▂▁▁ #365 of 431* 16% 2
coding · #376 of 469 · 18% of its score
Beats [LFM2.5 8B-A1B](lfm2-5-8b-a1b.md), [Qwen3-VL 2B Thinking](qwen3-vl-2b-thinking.md), [DeepSeek R1 Distill Qwen 7B](deepseek-r1-distill-qwen-7b.md) here, though they rank higher overall. Trails [Qwen2.5 1.5B Instruct](qwen2-5-1-5b.md), [WizardLM 13B V1.2](wizardlm-13b-v1-2.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
HumanEval (maker) 50.6% ▄·▄█·▄·●···█ beats 5/7 easy · avg 78% 9.1% sharp · at its level
MBPP (maker) 41.0% ▄···▄·▄▄●··█ beats 5/7 medium · avg 61% 4.5% at its level
CodeReviewQA: solution identifica… 14.2% ▃··▁▂●▃▂█▃▄▂ beats 11/47 medium · avg 51% 1.6% sharp · noisy
CodeReviewQA: automated code refi… 20.4% ▁▁▁▁▁▁▂●▃█▅▄ beats 12/47 hard · avg 32% 1.3% typical
CodeReviewQA: change localisation… 1.0% █●▁▁▂▂▂▃▅▅▅▇ beats 9/47 hard · avg 33% 1.1% sharp · noisy
CodeReviewQA: change type recogni… 69.0% ▁▁▁······▁●█ beats 18/47 easy · avg 67% 0.1% blunt · at its level · noisy
chat & writing · #348 of 402 · 10% of its score
Beats [Code Llama 70B Instruct](codellama-70b-instruct.md), [Gemma 4 E2B](gemma-4-e2b.md), [QwQ-32B-Preview](qwq-32b-preview.md) here, though they rank higher overall. Trails [Gemma 3 270M IT](gemma-3-270m.md), [Granite 4.0 H 350M](granite-4-0-h-350m.md), [LFM2 350M](lfm2-350m.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
LMArena (text) 1156.7 ▁▁▁▂●▃▃▆▆▇█▃ beats 45/385 avg 1260 7.4% blunt
IFEval (aggregation unspecified) … 46.1% ▄····█·●··██ beats 3/7 easy · avg 65% 2.8% at its level
math · #365 of 431 · 16% of its score
Beats [GPT-3.5 Turbo (0125)](gpt-3-5-turbo-0125.md), [DBRX Instruct](dbrx-instruct.md), [Command R Plus (04-2024)](command-r-plus-04-2024.md) here, though they rank higher overall. Trails [OPT-1.3B](opt-1-3b.md), [OPT-2.7B](opt-2-7b.md), [Gemma 2B (base)](gemma-2b-base.md) here, though they rank lower overall.
test result vs the others in its run difficulty counts why
GSM8K (maker) 59.7% █▄····▄··●▄▄ beats 5/7 easy · avg 80% 8.2% sharp · at its level
MATH (subset unspecified) (maker) 23.7% ▄···▄▄▄●··█▄ beats 4/7 medium · avg 49% 7.5% at its level
other tests (12)
test result vs the others in its run difficulty counts why
BIG-Bench Hard (maker) 43.7% ▃▃···▃··●··▅ beats 4/7 medium · avg 63% 14% sharp
MMLU (maker) 56.0% ▄····▄·▄●█·█ beats 3/7 easy · avg 76% 8.5% at its level
GPQA (subset unspecified) (maker) 28.6% ▄·▄··▄··●█·█ beats 3/7 hard · avg 37% 8.0% sharp · too hard for it
MMLU-Pro (maker) 27.9% ▄·····█··●▄█ beats 4/7 medium · avg 44% 5.9% typical
WinoGrande (maker) 71.9% █···▄▄▄●···█ beats 5/7 medium · avg 81% 5.3% at its level
HellaSwag (maker) 76.8% ▄·▄··█▄·●·▄▄ beats 5/7 easy · avg 85% 3.5% at its level
OpenBookQA (maker) 43.2% ▄▄··▄··▄·●▄▄ beats 5/7 medium · avg 55% 3.3% typical
ARC Challenge (maker) 54.4% ▄····█··▄●▄█ beats 4/7 medium · avg 72% 3.2% sharp · at its level
BoolQ (maker) 84.9% ▄▄····▄▄·●·█ beats 5/7 easy · avg 88% 2.2% at its level
EuroEval Life in the UK (English … 71.3% ▁▁▁▁▁▁▁▂●█▄▃ beats 40/146 easy · avg 82% 1.2% at its level · noisy
Vectara hallucination leaderboard… 91.2% ▁▁▂▃●▄█▆▃▂▂▁ beats 28/140 very easy · avg 94% 0.7% blunt · noisy
Social IQA (maker) 66.4% ▃·····▃··█▃● beats 6/7 medium · avg 66% 0.6% blunt · at its level
source results share of the score
IBM Granite 3.0 Language Models technic… 15 86% ▇▇▇▇▇▇▇▇▇▇
LMArena leaderboard dataset 1 7% ▇
CodeReviewQA: The Code Review Comprehen… 4 4% ▇
EuroEval leaderboards 1 1% ▇
Vectara hallucination leaderboard 1 1% ▇
prior: pull toward the middle 0% ▇
weighted average = 40.7
Licences: IBM Granite 3.0 Language Models technical report ([Apache-2.0](https://github.com/ibm-granite/granite-3.0-language-models/blob/5960b7fb3308588bf5f86348278ec9d6015ab647/LICENSE)) · LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) · CodeReviewQA: The Code Review Comprehension Assessment for Large Language Models ([CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)) · EuroEval leaderboards ([MIT](https://github.com/EuroEval/leaderboards/blob/1a77f1d2f8ba1ba05ab73287fb090d76bf41b2e0/LICENSE)) · Vectara hallucination leaderboard ([Apache-2.0](https://www.apache.org/licenses/LICENSE-2.0))
## How to run it
### 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 3.0 2B Instruct (Q4_K_M, needs ~1.6 GB, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/granite-3.0-2b-instruct-GGUF/tree/0f35cb534c61d4f1ea9a8e266efc522db70dc2fa))
```sh
llama cli -hf lmstudio-community/granite-3.0-2b-instruct-GGUF:Q4_K_M
```
3. Or start a local server and open the address it prints (http://localhost:9931)
```sh
llama serve -hf lmstudio-community/granite-3.0-2b-instruct-GGUF:Q4_K_M
```
Other versions: Q8_0 (~2.8 GB) `llama cli -hf lmstudio-community/granite-3.0-2b-instruct-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 3.0 2B Instruct (Q4_K_M, needs ~1.6 GB, from [huggingface.co/lmstudio-community](https://huggingface.co/lmstudio-community/granite-3.0-2b-instruct-GGUF/tree/0f35cb534c61d4f1ea9a8e266efc522db70dc2fa))
```sh
lms get https://huggingface.co/lmstudio-community/granite-3.0-2b-instruct-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 (~2.8 GB) `lms get https://huggingface.co/lmstudio-community/granite-3.0-2b-instruct-GGUF@Q8_0 -y` · 4bit (~1.8 GB, Mac only) `lms get https://huggingface.co/ekryski/granite-3.0-2b-instruct-4bit -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 3.0 2B Instruct (Q4_K_M, needs ~2.08 GB, from [ollama.com](https://ollama.com/library/granite3-dense/tags))
```sh
ollama run granite3-dense:2b-instruct-q4_K_M
```
Other versions: Q8_0 (~3.66 GB) `ollama run granite3-dense:2b-instruct-q8_0`
### 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 3.0 2B Instruct (4bit, needs ~1.8 GB, Mac only, from [huggingface.co/ekryski](https://huggingface.co/ekryski/granite-3.0-2b-instruct-4bit/tree/ffef3dffc738f5ec8ec984a1504202c9c1537859))
```sh
mlx_lm.chat --model ekryski/granite-3.0-2b-instruct-4bit
```
3. Or start an OpenAI-compatible server at http://localhost:8080/v1
```sh
mlx_lm.server --model ekryski/granite-3.0-2b-instruct-4bit
```
Model repository and documentation: [huggingface.co/ibm-granite](https://huggingface.co/ibm-granite/granite-3.0-2b-instruct)
[interactive version](https://whatmodel.sh/model/granite-3-0-2b) · [JSON](https://whatmodel.sh/api?format=json&all=1)