whatmodel.sh / Granite 3.0 3B-A800M Instruct
Granite 3.0 3B-A800M Instruct
↗ share
+ contribute
Granite 3.0 3B-A800M Instruct is an open-weights language model from IBM, released October 2024. Its published size is 3.3B parameters (0.8B active per token) and a 4k-token context window. Licence: apache-2.0.
Catalogue identity notes
IBM granite-3.0-3b-a800m-instruct (ibm-granite/granite-3.0-3b-a800m-instruct @51ced9b, created 2024-10-03, apache-2.0, 3,374,295,552 params; card '# Parameters 3.3B', '# Active Parameters 800M'). Lab rows from the Granite 3.0 paper Table 11 (MoE instruction models, IBM's harness, with the catalogued Llama-3.2 1B column) joining the existing Table-10 runs of that source; independent EuroEval rows.
Model information sources
Release date ↗ · Weights licence ↗
Card architecture table '# Parameters 3.3B' / '# Active Parameters 800M' (HF safetensors metadata 3,374,295,552).
$ how it compares
The LLMs ranked either side of Granite 3.0 3B-A800M Instruct, out of 828 with a score, then more from IBM.
| # | model | score | where | ||
|---|---|---|---|---|---|
| 697 |  Falcon 40B (base) | 35±1 | local | ||
| 698 |  StarCoder2 7B | 35±1 | local | ||
| 699 | Granite 3.0 3B-A800M Instruct | 35±1 | local | ||
| 700 |  DeepSeek LLM 7B Chat | 35±2 | local | ||
| 701 | StarCoder 15.5B | 35±1 | local | ||
| 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
coding▁▁▁▃▆██▇▆▄▃▁ #409 of 458 * · 16% of its score · 2 tests
Beats DeepSeek R1 Distill Qwen 7B, Qwen3.5 2B, DeepSeek R1 Distill Llama 8B here, though they rank higher overall. Trails Llama 3.2 1B Instruct, SmolLM2 1.7B Instruct, DeepSeek LLM 7B Chat here, though they rank lower overall.
chat & writing▁▁▁▃▅▄▅█▆▆▇▃ #348 of 393 * · 3.3% of its score · 1 test
Beats Gemma 4 E2B, Code Llama 70B Instruct, Phi-3-mini-4k-instruct (June 2024 update) here, though they rank higher overall. Trails Gemma 3 270M IT, SmolLM2 360M Instruct, Granite 4.0 H 350M here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| IFEval (aggregation unspecified) lab | 42.5% ✎ | ▄····█·▄··██ beats 2/7 | 3.3% |
math▁▂▃▅▆▅▄█▄▁▁▁ #361 of 398 * · 16% of its score · 2 tests
Beats Yi 34B Chat, Llama 3 8B Instruct, Command R (03-2024) here, though they rank higher overall. Trails Gemma 3 1B, Stable Code 3B, Qwen3 0.6B here, though they rank lower overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| GSM8K lab | 47.5% ✎ | █▄····▄··█▄▄ beats 3/7 | 8.2% |
| MATH (subset unspecified) lab | 19.9% ✎ | ▄···▄▄▄▄··█▄ beats 2/7 | 7.5% |
other tests11 not in a task · 65% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| BIG-Bench Hard lab | 37.7% ✎ | ▃▃···▃··█··▅ beats 2/7 | 13% |
| WinoGrande lab | 67.0% ✎ | █···▄▄▄▄···█ beats 2/7 | 10% |
| MMLU lab | 50.2% ✎ | ▄····▄·▄▄█·█ beats 2/7 | 8.0% |
| GPQA (subset unspecified) lab | 26.9% ✎ | ▄·▄··▄··▄█·█ beats 2/7 | 6.4% |
| HellaSwag lab | 71.9% ✎ | ▄·▄··█▄·▄·▄▄ beats 3/7 | 6.0% |
| BoolQ lab | 78.6% ✎ | ▄▄····▄▄·█·█ beats 2/7 | 5.8% |
| MMLU-Pro lab | 20.5% ✎ | ▄·····█··█▄█ beats 2/7 | 5.7% |
| ARC Challenge lab | 50.9% ✎ | ▄····█··▄▄▄█ beats 3/7 | 3.9% |
3 more tests
| test | result | vs the others in its run | counts |
|---|---|---|---|
| OpenBookQA lab | 40.8% ✎ | ▄▄··▄··▄·█▄▄ beats 3/7 | 3.7% |
| EuroEval Life in the UK (English knowledge MC), test split, 5-shot, accuracy | 55.8% ✎ | ▁▁▁▁▁▁▁▃▄█▄▃ beats 27/146 | 1.3% |
| Social IQA lab | 60.0% ✎ | ▃·····▃··█▃▅ beats 3/7 | 1.0% |
$ 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 3.0 3B-A800M Instruct · needs ~2.1 GB · from huggingface.co/lmstudio-communitychat in the terminal
llama cli -hf lmstudio-community/granite-3.0-3b-a800m-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-3.0-3b-a800m-instruct-GGUF:Q4_K_M
LM Studio
version
- Install LM Studio · Open the app once before using the lms command.
- Run Granite 3.0 3B-A800M Instruct · needs ~2.1 GB · from huggingface.co/lmstudio-community
- Or download it in the terminal
lms get https://huggingface.co/lmstudio-community/granite-3.0-3b-a800m-instruct-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) ↗
Linuxcurl -fsSL https://ollama.com/install.sh | sh - Run Granite 3.0 3B-A800M Instruct · needs ~2.1 GB · from ollama.comchat in the terminal
ollama run granite3-moe:3b
Weights and documentation: huggingface.co/ibm-granite ↗
$ how the score was worked out · IBM Granite 3.0 Language Mode… 98%, EuroEval leaderboards (result… 1%
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 3.0 Language Models technical report (GitHub) · lab-reported | 15 | 98% |
| EuroEval leaderboards (results collection), English generative leaderboard | 1 | 1% |
| prior · pull toward the middle | 0% | |
| weighted average | = 34.7 |
$ 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