What are the hardware requirements for Galactica 30B?

Galactica 30B requires approximately 60.8 GB of VRAM and supports a context window of 2K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Galactica 30B and how large is it?

Galactica 30B is developed by facebook, a model with 30b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate Galactica 30B?

Galactica 30B is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Galactica 30B by facebook — VRAM 60.8GB, 2K context

Name: Galactica 30B
Author: facebook

Galactica 30B is an open-source language model by facebook. Features: 30b LLM, VRAM: 60.8GB, Context: 2K, License: cc-by-nc-4.0, LLM Explorer Score: 0.04.

Arxiv:1810.03993 Galactica Opt Pytorch Region:us Sharded

Model Card on HF 🤗: https://huggingface.co/facebook/galactica-30b

Galactica 30B Parameters and Internals

Model Type

Transformer, decoder-only

Use Cases

Areas:

science, research

Applications:

citation prediction, scientific QA, mathematical reasoning, summarization, document generation, molecular property prediction, entity extraction

Primary Use Cases:

scientific tasks, scientific tooling

Limitations:

hallucinations, popularity bias, bias on certain measures, prone to exhibit stereotypes and toxicity

Considerations:

The model continues to exhibit a popularity bias and hallucinations especially for less cited scientific concepts.

Additional Notes

GALACTICA models are designed to perform scientific tasks and are trained on a large-scale scientific corpus.

Training Details

Data Sources:

papers, textbooks, scientific websites, encyclopedias, reference material, knowledge bases

Data Volume:

106 billion tokens

Model Architecture:

Transformer based architecture in a decoder-only setup with a few modifications

LLM Name	Galactica 30B
Repository 🤗	https://huggingface.co/facebook/galactica-30b
Model Size	30b
Required VRAM	60.8 GB
Updated	2026-08-07
Maintainer	facebook
Model Type	opt
Model Files	9.8 GB: 1-of-7 9.9 GB: 2-of-7 9.9 GB: 3-of-7 9.9 GB: 4-of-7 9.9 GB: 5-of-7 9.9 GB: 6-of-7 1.5 GB: 7-of-7
Model Architecture	OPTForCausalLM
License	cc-by-nc-4.0
Context Length	2048
Model Max Length	2048
Transformers Version	4.21.0.dev0
Tokenizer Class	PreTrainedTokenizerFast
Vocabulary Size	50000
Torch Data Type	float16
Activation Function	gelu

Rank the Galactica 30B 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 Galactica 30B

Best Alternatives	Context / RAM	Downloads	Likes
Galpaca 30B MiniOrca	2K / 59.6 GB	48	1
Opt 30B	2K / 60.1 GB	14167	137
Galpaca 30B	2K / 60.8 GB	59	55
Opt Iml Max 30B	2K / 60.1 GB	1150	35
OPT 30B Erebus	2K / 36 GB	60	67
...alactica 30B Evol Instruct 70K	2K / 60.1 GB	57	23
Opt Iml 30B	2K / 60.1 GB	36	73
Dalio Io 30B Cp1500	2K / 121.7 GB	11	0
Galactica 30B	2K / GB	19	1
...h 1e 5 Lr Handwritten Complete	2K / 121.7 GB	18	0

Note: green Score (e.g. "73.2") means that the model is better than facebook/galactica-30b.