whatmodel.sh / BLOOM 3B

BLOOM 3B

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

BLOOM 3B is an open-weights language model from BigScience, released July 2022. Its published size is 3B parameters and a 2k-token context window. Licence: bigscience-bloom-rail-1.0.

Catalogue identity notes

BigScience BLOOM pretrained base model (not BLOOMZ).

17.7±0.9#696 of 719 LLMsfrom 22 benchmark results across 5 sources

$ how it compares

The LLMs ranked either side of BLOOM 3B, out of 828 with a score, then more from BigScience.

#modelscorewhereRAMreleased
801GPT-3 babbage (API base model) OpenAI18±1comparison only2020-06
802Pythia 1.4B EleutherAI18±2local~2.4GB2023-02
803BLOOM 3B BigScience18±1local~4GB2022-07
804FastChat-T5 3B LMSYS17±2local~8.5GB2023-04
805Dolly v2 12B Databricks17±2comparison only2023-04
more from BigScience
790BLOOM 7.1B BigScience21±1local~7.1GB2022-07
810BLOOM 1.7B BigScience14±1local~3.1GB2022-07
814BLOOM 1.1B BigScience11±1local~2.6GB2022-07

see all LLMs · find the ones that fit your machine

$ what it's good at

coding█···█··█··██ #2 of 5 * · 11% of its score · 1 test
testresultvs the others in its rundifficultycountswhy
HumanEval lab6.5% ✎▄···▄·▄·█··▄ beats 3/5easy · avg 78%▇▇▇▇ 11%sharp · too hard for it
other tests15 not in a task · 89% of its score
testresultvs the others in its rundifficultycountswhy
ARC Challenge +130.4% ✎▁·▁··▃▁▇▂█▆▃ beats 2/26easy · avg 85%▇▇▇▇▇▇▇ 19%sharp · too hard for it · several runs
ARC Easy +153.3% ✎▁▁·▃▁▄▅▄█▅▁▃ beats 2/26very easy · avg 90%▇▇▇▇▇ 14%at its level · several runs
SciQ +189.1% ✎▁·▁·▁·▃▅▃█▄▁ beats 4/26very easy · avg 98%▇▇▇▇▇ 13%several runs
WinoGrande +158.5% ✎▁····▁··▃▂█▄ beats 3/26easy · avg 89%▇▇▇▇▇ 11%several runs
PIQA +170.6% ✎▁·▁··▃▁▃█▅█▂ beats 2/26easy · avg 88%▇▇▇ 7.8%blunt · at its level · several runs
LAMBADA (OpenAI version), last-word accuracy +151.7% ✎▁··▃▁▁▁▂▆█▅▃ beats 1/26very easy · avg 89%▇▇ 6.1%blunt · at its level · several runs
BoolQ61.7% ✎▂▃▃▅▆▂▅▆█▂▂▂ beats 3/26easy · avg 91%▇▇ 6.1%typical
HellaSwag54.5% ✎▃···▃▇▅█▇▃▁▁ beats 2/26very easy · avg 92%▇▇ 5.1%at its level
7 more tests
testresultvs the others in its rundifficultycountswhy
Natural Questions (closed book)4.8% ✎▁·▁·▁▃▁▂·▅▂█ beats 5/28medium · avg 58%▇ 3.4%too hard for it
Social IQA40.2% ✎▁···▁▃█▁▁▁·▁ beats 2/26medium · avg 59%▇ 1.4%blunt
OpenBookQA21.8% ✎█·▁▁▃·▁▃▅▇▃▄ beats 3/26medium · avg 53%▇ 1.1%too hard for it
TriviaQA17.9% ✎▁··▁█▃▃▁▄▅▄█ beats 10/28easy · avg 79%▇ 0.7%noisy
Open PT-LLM Leaderboard average (9 Portuguese tasks)33.5% ✎▃▃▁▁▁▁▁▁▂█▅▁ beats 38/276medium · avg 73%▇ 0.5%noisy
TruthfulQA MC240.6% ✎▅▃█▆▁▂▅···▁▁ beats 10/31medium · avg 64%▇ 0.5%blunt · at its level
HaluEval QA hallucination detection (accuracy)50.4% ✎█▃▁·▁▁·▁▁▁▁▁ beats 16/27medium · avg 70%▇ 0.1%sharp · too hard for it · noisy

$ where to run it

llama.cpp · recommended

version

  1. 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 ↗

  2. Run BLOOM 3B · needs ~2.4 GB · from huggingface.co/mradermacher
    chat in the terminal
    llama cli -hf mradermacher/bloom-3b-GGUF:Q4_K_M
  3. Or start a local server and open the address it prints (http://localhost:9931)
    llama serve -hf mradermacher/bloom-3b-GGUF:Q4_K_M

LM Studio

version

  1. Install LM Studio · Open the app once before using the lms command.

    download it (macOS / Windows / Linux) ↗

  2. Run BLOOM 3B · needs ~2.4 GB · from huggingface.co/mradermacher

    open in LM Studio ↗

  3. Or download it in the terminal
    lms get https://huggingface.co/mradermacher/bloom-3b-GGUF@Q4_K_M -y
  4. Then run lms ls to see its name, and lms chat followed by that name

Ollama

version

  1. Install Ollama

    download it (macOS / Windows) ↗

    Linux
    curl -fsSL https://ollama.com/install.sh | sh
  2. Run BLOOM 3B · needs ~2.4 GB · from huggingface.co/mradermacher
    chat in the terminal
    ollama run hf.co/mradermacher/bloom-3b-GGUF:Q4_K_M

Other ways to run it

  1. Run it as published · needs ~6.01 GB · from huggingface.co/bigscience (official)
    uv run --with torch --with 'transformers<5' python -c "from transformers import pipeline; print(pipeline('text-generation', model='bigscience/bloom-3b')('Once upon a time', max_new_tokens=40)[0]['generated_text'])"

Official weights and documentation: huggingface.co/bigscience ↗

$ how the score was worked out · Stability AI — original Stabl… 54%, Pythia: A Suite for Analyzing… 30%, BLOOM: A 176B-Parameter Open-… 11% 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

sourceresultsshare of the scorelicence
Stability AI — original StableLM zero-shot evaluation outputs10▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 54%Apache-2.0
Pythia: A Suite for Analyzing Large Language Models Across Training and Scaling — authors' own runs6▇▇▇▇▇▇▇▇▇ 30%CC BY-SA 4.0
BLOOM: A 176B-Parameter Open-Access Multilingual Language Model — authors' own runs · lab-reported1▇▇▇ 11%CC BY 4.0
Hallucinations Leaderboard results (HF dataset hallucinations-leaderboard/results)4▇ 5%Apache-2.0
Open Portuguese LLM Leaderboard (archived)1▇ 1%Apache-2.0
prior · pull toward the middle▇ 0%
weighted average= 17.7

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit 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 · sources and licences · json