whatmodel.sh / Granite 3.1 2B Instruct
Granite 3.1 2B Instruct
↗ share
+ contribute
Granite 3.1 2B Instruct is an open-weights language model from IBM, released December 2024. Its published size is 2.53B parameters and a 128k-token context window. Licence: apache-2.0.
Catalogue identity notes
Arena entry 'granite-3.1-2b-instruct' (text). LMArena's leaderboard table (arena-leaderboard space @6f2eeeb) labels 'granite-3.1-2b-instruct' as 'Granite-3.1-2B-Instruct' (IBM, Apache 2.0), linking https://huggingface.co/ibm-granite/granite-3.1-2b-instruct. Maker repo ibm-granite/granite-3.1-2b-instruct @bbc2aed: licence tag apache-2.0. 2,533,531,648 parameters from the Hugging Face safetensors metadata (dense, so active = total). Released 2024-12: month the maker's Hugging Face repo was created (createdAt 2024-12-06).
$ how it compares
The LLMs ranked either side of Granite 3.1 2B Instruct, out of 828 with a score, then more from IBM.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 678 |  Zephyr 7B alpha | 36±1 | local | ||
| 679 |  Vicuna 13B v1.5 | 36±2 | local | ||
| 680 | Granite 3.1 2B Instruct | 36±1 | local | ||
| 681 |  EXAONE 3.5 2.4B Instruct | 36±2 | local | ||
| 682 |  LLaMA 33B | 36±1 | comparison only | ||
| more from IBM | |||||
| 192 | Granite 4.2 30B | 63±1 | api $0.28/M +local | ||
| 239 | Granite 4.2 8B | 60±1 | api $0.11/M +local | ||
| 276 | Granite 4.2 3B | 57±1 | api $0.052/M +local | ||
$ what it's good at
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #325 of 393 · 17% of its score · 3 tests
Beats OpenThinker3 7B, Gemma 4 E2B, Code Llama 70B Instruct here, though they rank higher overall. Trails LFM2 350M, Gemma 3 1B, Flan-T5 XXL here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| LMArena (text) | 1178.8 ✎ | ▁▁▁▂▃▃▃▆▆▇█▃ beats 55/379 | 8.1% |
| IFEval (aggregation unspecified) lab +1 | 63.6% ✎ | ▄▄·▄······▄█ beats 1/5 | 5.8% |
| Arena-Hard (version/judge unspecified) lab | 23.3% ✎ | █·▄··▄·····█ beats 0/5 | 3.2% |
math▁▂▃▅▆▅▄█▄▁▁▁ #355 of 398 · 31% of its score · 3 tests
Beats Yi 34B Chat, Llama 3 8B Instruct, Command R (03-2024) here, though they rank higher overall. Trails Gemma 3 1B, Qwen3 0.6B, Tulu 2 DPO 7B here, though they rank lower overall.
other tests9 not in a task · 52% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| MMLU lab +1 | 57.1% ✎ | ▄█········▄█ beats 1/5 | 10% |
| BIG-Bench Hard lab | 61.8% ✎ | █·▄·····▄▄·▄ beats 1/5 | 8.9% |
| ARC Challenge lab | 54.6% ✎ | ▆▁▁▂▂▅▇█▄▃▃▂ beats 64/97 | 8.8% |
| CPTUB average (SpeakLeash, 0-5 judge scale) | 38.9% ✎ | ▁▃▂▂▃█▃▆▅▄▂▃ beats 7/95 | 6.0% |
| DROP lab | 21.0% ✎ | ▃▃·····▃···█ beats 0/5 | 5.6% |
| WinoGrande lab | 67.5% ✎ | ▁···▂▁▁▅█▇▃▁ beats 52/87 | 5.5% |
| HellaSwag lab | 75.1% ✎ | ▁▁▁▁▂▂▃▇█▆▅▃ beats 55/85 | 3.9% |
| PopQA lab | 20.5% ✎ | ▄·█······▄·█ beats 1/5 | 2.4% |
1 more test
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Vectara hallucination leaderboard, 2025 dataset (HHEM-2.3, factual consistency rate) | 84.3% ✎ | ▁▁▃▃▄▄█▆▃▂▂▁ beats 13/140 | 0.9% |
$ where to run it
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 | shWindowsirm https://llama.app/install.ps1 | iex - Run Granite 3.1 2B Instruct · needs ~1.5 GB · from huggingface.co/lmstudio-communitychat in the terminal
llama cli -hf lmstudio-community/granite-3.1-2b-instruct-GGUF:Q4_K_M - Or start a local server and open the address it prints (http://localhost:9931)
llama serve -hf lmstudio-community/granite-3.1-2b-instruct-GGUF:Q4_K_M
LM Studio
version
- Install LM Studio · Open the app once before using the lms command.
- Run Granite 3.1 2B Instruct · needs ~1.5 GB · from huggingface.co/lmstudio-community
- Or download it in the terminal
lms get https://huggingface.co/lmstudio-community/granite-3.1-2b-instruct-GGUF@Q4_K_M -y - Then run lms ls to see its name, and lms chat followed by that name
Ollama
version
- Install Ollama
download it (macOS / Windows) ↗
Linuxcurl -fsSL https://ollama.com/install.sh | sh - Run Granite 3.1 2B Instruct · needs ~2.02 GB · from ollama.comchat in the terminal
ollama run granite3.1-dense:2b-instruct-q4_K_M
Weights and documentation: huggingface.co/ibm-granite ↗
$ how the score was worked out · ibm-granite/granite-3.3-8b-in… 52%, ibm-granite/granite-3.1-3b-a8… 32%, LMArena leaderboard dataset 8% and 2 more
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 |
|---|---|---|
 ibm-granite/granite-3.3-8b-instruct model card · lab-reported | 9 | 52% |
| ibm-granite/granite-3.1-3b-a800m-instruct model card · lab-reported | 6 | 32% |
 LMArena leaderboard dataset | 1 | 8% |
| SpeakLeash CPTUB (Complex Polish Text Understanding Benchmark) | 1 | 6% |
 Vectara hallucination leaderboard (vectara/hallucination-leaderboard repository) | 1 | 1% |
| prior · pull toward the middle | 0% | |
| weighted average | = 36.2 |
$ improve this page
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