What are the hardware requirements for Llama 4 Maverick 17B 128E Instruct FP8?

Llama 4 Maverick 17B 128E Instruct FP8 requires approximately 210 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Llama 4 Maverick 17B 128E Instruct FP8 and how large is it?

Llama 4 Maverick 17B 128E Instruct FP8 is developed by unsloth, a model with 17b 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 Llama 4 Maverick 17B 128E Instruct FP8?

Llama 4 Maverick 17B 128E Instruct FP8 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Llama 4 Maverick 17B 128E Instruct FP8 by unsloth — VRAM 210GB

Name: Llama 4 Maverick 17B 128E Instruct FP8
Author: unsloth

Llama 4 Maverick 17B 128E Instruct FP8 is an open-source language model by unsloth. Features: 17b LLM, VRAM: 210GB, License: other, Instruction-Based, LLM Explorer Score: 0.19.

Arxiv:2204.05149 Ar Base model:meta-llama/llama-4-... Base model:quantized:meta-llam... Compressed-tensors Conversational De Deploy:azure En Endpoints compatible Es Facebook Fr Hi Id Image-text-to-text Instruct It Llama Llama-4 Llama4 Meta Pt Pytorch Region:us Safetensors Sharded Tensorflow Th Tl Vi

Model Card on HF 🤗: https://huggingface.co/unsloth/Llama-4-Maverick-17B-128E-Instruct-FP8

Llama 4 Maverick 17B 128E Instruct FP8 Parameters and Internals

LLM Name	Llama 4 Maverick 17B 128E Instruct FP8
Repository 🤗	https://huggingface.co/unsloth/Llama-4-Maverick-17B-128E-Instruct-FP8
Base Model(s)	meta-llama/Llama-4-Maverick-17B-128E-Instruct meta-llama/Llama-4-Maverick-17B-128E-Instruct
Model Size	17b
Required VRAM	210 GB
Updated	2026-08-10
Maintainer	unsloth
Model Type	llama4
Instruction-Based	Yes
Model Files	5.0 GB: 1-of-84 5.0 GB: 2-of-84 5.0 GB: 3-of-84 5.0 GB: 4-of-84 5.0 GB: 5-of-84 5.0 GB: 6-of-84 5.0 GB: 7-of-84 5.0 GB: 8-of-84 5.0 GB: 9-of-84 5.0 GB: 10-of-84 5.0 GB: 11-of-84 5.0 GB: 12-of-84 5.0 GB: 13-of-84 5.0 GB: 14-of-84 5.0 GB: 15-of-84 5.0 GB: 16-of-84 5.0 GB: 17-of-84 5.0 GB: 18-of-84 5.0 GB: 19-of-84 5.0 GB: 20-of-84 5.0 GB: 21-of-84 5.0 GB: 22-of-84 5.0 GB: 23-of-84 5.0 GB: 24-of-84 5.0 GB: 25-of-84 5.0 GB: 26-of-84 5.0 GB: 27-of-84 5.0 GB: 28-of-84 5.0 GB: 29-of-84 5.0 GB: 30-of-84 5.0 GB: 31-of-84 5.0 GB: 32-of-84 5.0 GB: 33-of-84 5.0 GB: 34-of-84 5.0 GB: 35-of-84 5.0 GB: 36-of-84 5.0 GB: 37-of-84 5.0 GB: 38-of-84 5.0 GB: 39-of-84 5.0 GB: 40-of-84 5.0 GB: 41-of-84 5.0 GB: 42-of-84
Supported Languages	ar de en es fr hi id it pt th tl vi
Model Architecture	Llama4ForConditionalGeneration
License	other
Transformers Version	4.52.3
Torch Data Type	bfloat16

Rank the Llama 4 Maverick 17B 128E Instruct FP8 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 Llama 4 Maverick 17B 128E Instruct FP8

Best Alternatives	Context / RAM	Downloads	Likes
Llama 4 Scout 17B 16E Instruct	0K / 188.7 GB	3586	57
...a 4 Maverick 17B 128E Instruct	0K / 631.8 GB	3508	8
...Scout 17B 16E Instruct Unsloth	0K / 197.5 GB	23	17
...Maverick 17B 16E Instruct 4bit	0K / 146.3 GB	419	7
... 4 Scout 17B 16E Instruct 6bit	0K / 88.8 GB	333	5
...averick 17B 128E Instruct 6bit	0K / 201.1 GB	125	1
... 4 Scout 17B 16E Instruct 4bit	0K / 61 GB	1203	12
... 16E Instruct Unsloth Bnb 4bit	0K / 60.6 GB	205	80
... 4 Scout 17B 16E Instruct 8bit	0K / 114.7 GB	588	4
...averick 17B 128E Instruct 4bit	0K / 140.8 GB	204	2

Note: green Score (e.g. "73.2") means that the model is better than unsloth/Llama-4-Maverick-17B-128E-Instruct-FP8.