What are the hardware requirements for Garrulus AWQ?

Garrulus AWQ requires approximately 4.2 GB of VRAM and supports a context window of 32K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Garrulus AWQ and how large is it?

Garrulus AWQ is developed by TheBloke, a model with 7.2b 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 Garrulus AWQ?

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

Garrulus AWQ by TheBloke — VRAM 4.2GB, 32K context

Name: Garrulus AWQ
Author: TheBloke

Garrulus AWQ is an open-source language model by TheBloke. Features: 7.2b LLM, VRAM: 4.2GB, Context: 32K, License: apache-2.0, Quantized, LLM Explorer Score: 0.11.

4-bit 7b Awq Base model:quantized:udkai/gar... Base model:udkai/garrulus Dataset:hromi/winograd dpo bas... Dpo Mistral Mlabonne/neuralmarcoro14-7b Quantized Region:us Safetensors Winograd

Model Card on HF 🤗: https://huggingface.co/TheBloke/Garrulus-AWQ

Garrulus AWQ Parameters and Internals

Model Type

mistral

Additional Notes

This model was initially named 'ContaminatedWine' but was later named 'Garrulus' after an encounter with a eurasian jay. The model shows improved performance not only on Winogrande but also on TruthfulQA, HellaSwag, and ARC challenge metrics due to DPO contamination with Winogrande.

Training Details

Data Sources:

mlabonne/NeuralMarcoro14-7B, hromi/winograd_dpo_basic

Methodology:

Intentionally contaminated with two epochs of direct preference optimization (DPO) with a slightly modified Winogrande dataset.

Training Time:

Few minutes on a A40

Hardware Used:

Hardware of Berlin University of the Arts (UDK-berlin)

LLM Name	Garrulus AWQ
Repository 🤗	https://huggingface.co/TheBloke/Garrulus-AWQ
Model Name	Garrulus
Model Creator	UDK dot AI
Base Model(s)	Garrulus udkai/Garrulus
Model Size	7.2b
Required VRAM	4.2 GB
Updated	2026-08-08
Maintainer	TheBloke
Model Type	mistral
Model Files	4.2 GB
AWQ Quantization	Yes
Quantization Type	awq
Model Architecture	MistralForCausalLM
License	apache-2.0
Context Length	32768
Model Max Length	32768
Transformers Version	4.37.0.dev0
Tokenizer Class	LlamaTokenizer
Padding Token	<unk>
Vocabulary Size	32000
Torch Data Type	float16

Rank the Garrulus AWQ 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 Garrulus AWQ

Best Alternatives	Context / RAM	Downloads	Likes
Feedback Wizard 4bit AWQ	32K / 4.2 GB	6	0
Ryze AWQ 4bit	32K / 4.2 GB	5	0
...V3 Plot Chat Roleplay Chat AWQ	32K / 4.2 GB	7	0
...lot Engine Step Json Short AWQ	32K / 4.2 GB	5	0
MistralLite AWQ	32K / 4.2 GB	27	6
LemonadeRP 4.5.3 AWQ	32K / 4.2 GB	5	0
Functionary Small V2.4 AWQ	32K / 4.2 GB	1	1
Hc Mistral Alpaca Merged AWQ	32K / 4.2 GB	5	0
HamSter 0.1 AWQ	32K / 4.2 GB	5	2
Phoenix AWQ	32K / 4.2 GB	12	1

Note: green Score (e.g. "73.2") means that the model is better than TheBloke/Garrulus-AWQ.