What are the hardware requirements for Nucleus 22B Token 500B?

Nucleus 22B Token 500B requires approximately 87.5 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 Nucleus 22B Token 500B and how large is it?

Nucleus 22B Token 500B is developed by NucleusAI, a model with 22b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Nucleus 22B Token 500B perform on standard benchmarks?

Nucleus 22B Token 500B has the following published scores: MMLU 30.11. Compare against reference models on this page or on the LLM Explorer leaderboards.

Nucleus 22B Token 500B by NucleusAI — VRAM 87.5GB, 2K context

Name: Nucleus 22B Token 500B
Author: NucleusAI

Nucleus 22B Token 500B is an open-source language model by NucleusAI. Features: 22b LLM, VRAM: 87.5GB, Context: 2K, License: mit, LLM Explorer Score: 0.13, Arc: 40.7, HellaSwag: 69.4, MMLU: 30.1, GSM8K: 1.

En Endpoints compatible Llama Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/NucleusAI/nucleus-22B-token-500B

Nucleus 22B Token 500B Parameters and Internals

Model Type

Causal decoder-only

Use Cases

Primary Use Cases:

Research on large language models, Foundation for further specialization and finetuning for specific usecases like summarization, text generation, chatbot

Considerations:

Model should be finetuned for specific tasks of interest. Precautions are needed for any production use.

Additional Notes

This is a raw, pretrained model, which should be further finetuned for most use cases. Nucleus-22B-token-500B is trained on English data and won't generalize to other languages. It may carry online stereotypes and biases.

Supported Languages

English (NLP)

Training Details

Data Sources:

RefinedWeb-English, Books, Code, Technical, Math

Data Volume:

500B tokens

Methodology:

Tokenized with the tokenizer similar to Llama-7B

Context Length:

2048

Training Time:

Two weeks in early August 2023

Hardware Used:

256 A100 80GB GPUs

LLM Name	Nucleus 22B Token 500B
Repository 🤗	https://huggingface.co/NucleusAI/nucleus-22B-token-500B
Model Size	22b
Required VRAM	87.5 GB
Updated	2026-08-07
Maintainer	NucleusAI
Model Type	llama
Model Files	9.9 GB: 1-of-9 9.7 GB: 2-of-9 9.8 GB: 3-of-9 10.0 GB: 4-of-9 9.8 GB: 5-of-9 9.9 GB: 6-of-9 9.9 GB: 7-of-9 9.8 GB: 8-of-9 8.7 GB: 9-of-9
Supported Languages	en
Model Architecture	LlamaForCausalLM
License	mit
Context Length	2048
Model Max Length	2048
Transformers Version	4.30.2
Tokenizer Class	LlamaTokenizer
Beginning of Sentence Token	<s>
End of Sentence Token	</s>
Unk Token	<unk>
Vocabulary Size	32000
Torch Data Type	float32

Rank the Nucleus 22B Token 500B 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

Quantized Models of the Nucleus 22B Token 500B

Model	Likes	Downloads	VRAM
Nucleus 22B Token 500B GGUF	3	345	9 GB
Nucleus 22B Token 500B GPTQ	2	7	11 GB
Nucleus 22B Token 500B AWQ	1	4	12 GB

Best Alternatives to Nucleus 22B Token 500B

Best Alternatives	Context / RAM	Downloads	Likes
EuroLLM 22B Instruct 2512	32K / 45.5 GB	3342	63
EuroLLM 22B 2512	32K / 45.5 GB	2766	19
Calm3 22B Chat	16K / 44.9 GB	878	85
Calm3 22B RP V2	16K / 44.9 GB	21	20
Calm3 22B RP V0.1	16K / 44.9 GB	10	0
Yousei 22B	4K / 44.5 GB	97	2
Llama2 22B Daydreamer V3	4K / 43.7 GB	177	11
Platypus 2 22B Relora	4K / 43.7 GB	93	1
Llama2 22B Blocktriangular	4K / 43.7 GB	368	4
Llama2 22B	4K / 43.7 GB	109	46

Note: green Score (e.g. "73.2") means that the model is better than NucleusAI/nucleus-22B-token-500B.