What are the hardware requirements for Llama3 2x8b MoE 41K Experiment1?

Llama3 2x8b MoE 41K Experiment1 requires approximately 27.3 GB of VRAM and supports a context window of 8K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Llama3 2x8b MoE 41K Experiment1 and how large is it?

Llama3 2x8b MoE 41K Experiment1 is developed by nisten, a model with 13.7b 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 Llama3 2x8b MoE 41K Experiment1?

Llama3 2x8b MoE 41K Experiment1 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Llama3 2x8b MoE 41K Experiment1 by nisten — VRAM 27.3GB, 8K context

Name: Llama3 2x8b MoE 41K Experiment1
Author: nisten

Llama3 2x8b MoE 41K Experiment1 is an open-source language model by nisten. Features: 13.7b LLM, VRAM: 27.3GB, Context: 8K, License: llama3, MoE, Instruction-Based, LLM Explorer Score: 0.12.

Base model:finetune:meta-llama... Base model:meta-llama/meta-lla... Conversational Dataset:argilla/distilabel-cap... En Endpoints compatible Instruct Mixtral Moe Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/nisten/llama3-2x8b-MoE-41k-experiment1

Llama3 2x8b MoE 41K Experiment1 Parameters and Internals

Model Type

moe

Additional Notes

Experimental mixture of expert model with just 2 experts based on Llama 3 Instruct.

Supported Languages

en (proficient)

Training Details

Data Sources:

argilla/distilabel-capybara-dpo-7k-binarized

Methodology:

Mixture of Expert Model with fine-tuning

Model Architecture:

Two expert model

LLM Name	Llama3 2x8b MoE 41K Experiment1
Repository 🤗	https://huggingface.co/nisten/llama3-2x8b-MoE-41k-experiment1
Base Model(s)	Meta Llama 3 8B Instruct meta-llama/Meta-Llama-3-8B-Instruct
Model Size	13.7b
Required VRAM	27.3 GB
Updated	2026-07-29
Maintainer	nisten
Model Type	mixtral
Instruction-Based	Yes
Model Files	5.0 GB: 1-of-6 4.9 GB: 2-of-6 5.0 GB: 3-of-6 5.0 GB: 4-of-6 4.9 GB: 5-of-6 2.5 GB: 6-of-6
Supported Languages	en
Model Architecture	MixtralForCausalLM
License	llama3
Context Length	8192
Model Max Length	8192
Transformers Version	4.41.0.dev0
Tokenizer Class	PreTrainedTokenizerFast
Padding Token	<\|end_of_text\|>
Vocabulary Size	128256
Torch Data Type	bfloat16

Rank the Llama3 2x8b MoE 41K Experiment1 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 Llama3 2x8b MoE 41K Experiment1

Best Alternatives	Context / RAM	Downloads	Likes
L3.1 Celestial Stone 2x8B	128K / 27.3 GB	312	3
...ma 3 2x8B Instruct MoE 64K Ctx	64K / 27.3 GB	12	4
Defne Llama3 2x8B	8K / 27.4 GB	15	6
Inixion 2x8B V2	8K / 27.4 GB	6	2
MoE Llama3 8bx2 Rag	8K / 27.3 GB	7	0
Inixion 2x8B	8K / 27.5 GB	9	1
Llama 3 Chatty 2x8B	8K / 27.3 GB	6	11
FinalFintetuning XVIII 2x8B	8K / 27.5 GB	11	4
Llama 3 Teal Instruct 2x8B MoE	8K / 27.3 GB	8	0
Llama 3 8Bx2 MoE DPO	8K / 27.4 GB	4	1

Note: green Score (e.g. "73.2") means that the model is better than nisten/llama3-2x8b-MoE-41k-experiment1.