What are the hardware requirements for Notux 8x7b V1.3.5bpw H6 EXL2?

Notux 8x7b V1.3.5bpw H6 EXL2 requires approximately 20.7 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 Notux 8x7b V1.3.5bpw H6 EXL2 and how large is it?

Notux 8x7b V1.3.5bpw H6 EXL2 is developed by LoneStriker. 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 Notux 8x7b V1.3.5bpw H6 EXL2?

Notux 8x7b V1.3.5bpw H6 EXL2 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Notux 8x7b V1.3.5bpw H6 EXL2 by LoneStriker — VRAM 20.7GB, 32K context

Name: Notux 8x7b V1.3.5bpw H6 EXL2
Author: LoneStriker

Notux 8x7b V1.3.5bpw H6 EXL2 is an open-source language model by LoneStriker. Features: LLM, VRAM: 20.7GB, Context: 32K, License: apache-2.0, MoE, Quantized, Instruction-Based, LLM Explorer Score: 0.1.

Base model:finetune:mistralai/... Base model:mistralai/mixtral-8... Conversational Dataset:argilla/ultrafeedback-... De Dpo En Endpoints compatible Es Exl2 Fr Instruct It Mixtral Moe Preference Quantized Region:us Rlaif Safetensors Sharded Tensorflow Ultrafeedback

Model Card on HF 🤗: https://huggingface.co/LoneStriker/notux-8x7b-v1-3.5bpw-h6-exl2

Notux 8x7b V1.3.5bpw H6 EXL2 Parameters and Internals

Model Type

Pretrained generative Sparse Mixture of Experts

Supported Languages

English (Full proficiency), Spanish (Full proficiency), Italian (Full proficiency), German (Full proficiency), French (Full proficiency)

Training Details

Data Sources:

argilla/ultrafeedback-binarized-preferences-cleaned

Methodology:

Direct Preference Optimization (DPO) and experimental preference tuning methods like distilled DPO (dDPO)

Training Time:

1 epoch (~10hr)

Hardware Used:

8 x H100 80GB hosted in runpod.io

LLM Name	Notux 8x7b V1.3.5bpw H6 EXL2
Repository 🤗	https://huggingface.co/LoneStriker/notux-8x7b-v1-3.5bpw-h6-exl2
Base Model(s)	mistralai/Mixtral-8x7B-Instruct-v0.1 mistralai/Mixtral-8x7B-Instruct-v0.1
Required VRAM	20.7 GB
Updated	2026-07-24
Maintainer	LoneStriker
Model Type	mixtral
Instruction-Based	Yes
Model Files	8.6 GB: 1-of-3 8.6 GB: 2-of-3 3.5 GB: 3-of-3
Supported Languages	en de es fr it
Quantization Type	exl2
Model Architecture	MixtralForCausalLM
License	apache-2.0
Context Length	32768
Model Max Length	32768
Transformers Version	4.36.0
Tokenizer Class	LlamaTokenizer
Padding Token	</s>
Vocabulary Size	32000
Torch Data Type	bfloat16

Rank the Notux 8x7b V1.3.5bpw H6 EXL2 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 Notux 8x7b V1.3.5bpw H6 EXL2

Best Alternatives	Context / RAM	Downloads	Likes
...M 2 8x22B Beige 5.0bpw H6 EXL2	64K / 88.5 GB	8	0
...M 2 8x22B Beige 4.0bpw H6 EXL2	64K / 70.8 GB	6	0
...M 2 8x22B Beige 2.4bpw H6 EXL2	64K / 42.7 GB	5	0
...M 2 8x22B Beige 3.0bpw H6 EXL2	64K / 53.2 GB	5	0
...B Instruct V0.1 8.0bpw H8 EXL2	64K / 120.2 GB	6	1
...8x22b Instruct Oh EXL2 2.25bpw	64K / 40.1 GB	5	1
...eryTour V2 8x7B 4.5bpw H6 EXL2	32K / 26.5 GB	5	2
...it MoE 2bitgs8 Metaoffload HQQ	32K / 24.1 GB	9	19
...hin 2.7 Mixtral 8x7b 8bpw EXL2	32K / 46.8 GB	5	2
... 4bit MoE 3bit Metaoffload HQQ	32K / 22.4 GB	12	13

Note: green Score (e.g. "73.2") means that the model is better than LoneStriker/notux-8x7b-v1-3.5bpw-h6-exl2.