What are the hardware requirements for Huginn 22B Prototype GPTQ?

Huginn 22B Prototype GPTQ requires approximately 12 GB of VRAM and supports a context window of 4K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Huginn 22B Prototype GPTQ and how large is it?

Huginn 22B Prototype GPTQ is developed by TheBloke, a model with 22b 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 Huginn 22B Prototype GPTQ?

Huginn 22B Prototype GPTQ is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Huginn 22B Prototype GPTQ by TheBloke — VRAM 12GB, 4K context

Name: Huginn 22B Prototype GPTQ
Author: TheBloke

Huginn 22B Prototype GPTQ is an open-source language model by TheBloke. Features: 22b LLM, VRAM: 12GB, Context: 4K, License: llama2, Quantized, LLM Explorer Score: 0.08.

4-bit Base model:quantized:the-face-... Base model:the-face-of-goonery... Gptq Llama Quantized Region:us Safetensors

Model Card on HF 🤗: https://huggingface.co/TheBloke/Huginn-22B-Prototype-GPTQ

Huginn 22B Prototype GPTQ Parameters and Internals

Model Type

llama

Additional Notes

Model is a prototype, merged but not finetuned, and uses Alpaca format or chat prompts.

Training Details

Methodology:

Merged with Huginn v3 using chargoddard's Frankenllama script; has not been finetuned.

Input Output

Input Format:

Alpaca format

LLM Name	Huginn 22B Prototype GPTQ
Repository 🤗	https://huggingface.co/TheBloke/Huginn-22B-Prototype-GPTQ
Model Name	Huginn 22B Prototype
Model Creator	Caleb Morgan
Base Model(s)	The-Face-Of-Goonery/Huginn-22b-Prototype The-Face-Of-Goonery/Huginn-22b-Prototype
Model Size	22b
Required VRAM	12 GB
Updated	2026-07-15
Maintainer	TheBloke
Model Type	llama
Model Files	12.0 GB
GPTQ Quantization	Yes
Quantization Type	gptq
Model Architecture	LlamaForCausalLM
License	llama2
Context Length	4096
Model Max Length	4096
Transformers Version	4.31.0
Tokenizer Class	LlamaTokenizer
Beginning of Sentence Token	<s>
End of Sentence Token	</s>
Unk Token	<unk>
Vocabulary Size	32000
Torch Data Type	float16

Rank the Huginn 22B Prototype GPTQ 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 Huginn 22B Prototype GPTQ

Best Alternatives	Context / RAM	Downloads	Likes
Llama2 22B Daydreamer V3 GPT	4K / 12 GB	7	0
Llama2 22B Daydreamer V3 GPTQ	4K / 12 GB	10	6
Llama2 22B Daydreamer V2 GPTQ	4K / 12 GB	8	8
Ypotryll 22B Gptq	4K / 12 GB	9	1
Llama2 22B GPLATTY GPT	4K / 12 GB	6	0
Llama2 22B GPLATTY GPTQ	4K / 12 GB	4	4
Nucleus 22B Token 500B GPTQ	2K / 11.6 GB	7	2
...LLM 22B Instruct 2512 Mlx 4bit	32K / 12.7 GB	127	3
...LLM 22B Instruct 2512 Mlx 8bit	32K / 24.1 GB	53	1
Dendrite 22Bchk2 F16	4K / 43.7 GB	7	1

Note: green Score (e.g. "73.2") means that the model is better than TheBloke/Huginn-22B-Prototype-GPTQ.