What are the hardware requirements for Bloom 560M?

Bloom 560M requires approximately 1.1 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Bloom 560M and how large is it?

Bloom 560M is developed by bigscience, a model with 560m parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Bloom 560M perform on standard benchmarks?

Bloom 560M has the following published scores: MMLU 24.22. Compare against reference models on this page or on the LLM Explorer leaderboards.

Bloom 560M by bigscience — VRAM 1.1GB

Name: Bloom 560M
Author: bigscience

Bloom 560M is an open-source language model by bigscience. Features: 560m LLM, VRAM: 1.1GB, License: bigscience-bloom-rail-1.0, LLM Explorer Score: 0.16, Arc: 24.7, HellaSwag: 37.2, MMLU: 24.2, GSM8K: 0.3.

Arxiv:1909.08053 Arxiv:2108.12409 Arxiv:2110.02861 Ak Ar As Bloom Bm Bn Ca Code Deploy:azure En Endpoints compatible Es Eu Fon Fr Gu Hi Id Ig Jax Ki Kn Lg Ln Ml Mr Ne Nso Ny Onnx Or Pa Pt Pytorch Region:us Rn Rw Safetensors Sn St Sw Ta Te Tn Ts Tum Tw Ur Vi Wo Xh Yo Zh Zhs Zht Zu

Model Card on HF 🤗: https://huggingface.co/bigscience/bloom-560m

Bloom 560M Parameters and Internals

Model Type

text-generation

Supported Languages

ak (Akan), ar (Arabic), as (Assamese), bm (Bambara), bn (Bengali), ca (Catalan), code (Programming Languages), en (English), es (Spanish), eu (Basque), fon (Fon), fr (French), gu (Gujarati), hi (Hindi), id (Indonesian), ig (Igbo), ki (Kikuyu), kn (Kannada), lg (Luganda), ln (Lingala), ml (Malayalam), mr (Marathi), ne (Nepali), nso (Northern Sotho), ny (Nyanja), or (Oriya), pa (Punjabi), pt (Portuguese), rn (Rundi), rw (Kinyarwanda), sn (Shona), st (Southern Sotho), sw (Swahili), ta (Tamil), te (Telugu), tn (Tswana), ts (Tsonga), tum (Tumbuka), tw (Twi), ur (Urdu), vi (Vietnamese), wo (Wolof), xh (Xhosa), yo (Yoruba), zh (Chinese), zhs (Simplified Chinese), zht (Traditional Chinese), zu (Zulu)

Training Details

Data Sources:

45 natural languages, 12 programming languages

Data Volume:

1.5TB of pre-processed text

Methodology:

Byte-level Byte Pair Encoding (BPE) algorithm

Context Length:

2048

Training Time:

March 11, 2022, to July 5, 2022

Hardware Used:

384 A100 80GB GPUs

Model Architecture:

Modified from Megatron-LM GPT2

LLM Name	Bloom 560M
Repository 🤗	https://huggingface.co/bigscience/bloom-560m
Model Size	560m
Required VRAM	1.1 GB
Updated	2026-07-16
Maintainer	bigscience
Model Type	bloom
Model Files	1.1 GB 1.1 GB
Supported Languages	ak ar as bm bn ca code en es eu fr gu hi id ig ki kn lg ln ml mr ne ny or pa pt rn rw sn st sw ta te tn ts tw ur vi wo xh yo zh zu
Model Architecture	BloomForCausalLM
License	bigscience-bloom-rail-1.0
Transformers Version	4.20.0
Tokenizer Class	BloomTokenizerFast
Padding Token	<pad>
Vocabulary Size	250880

Rank the Bloom 560M capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to Bloom 560M

Best Alternatives	Context / RAM	Downloads	Likes
Bloomz 560M	0K / 1.1 GB	1150613	137
Train Test Bloom560	0K / 2.2 GB	14	0
Bloomz 560M Sft Chat	0K / 1.1 GB	152	10
Bloom 560M RLHF V2	0K / 1.1 GB	504	3
Bloom 560M RLHF	0K / 1.1 GB	499	1
Promt Generator	0K / 2.2 GB	930	43
Train Test	0K / 2.2 GB	30	0
Product Description Fr	0K / 2.2 GB	7	0
Guitester	0K / 2.2 GB	5	0
ModeloAJustadoBloom1	0K / 2.2 GB	10	0

Note: green Score (e.g. "73.2") means that the model is better than bigscience/bloom-560m.