What are the hardware requirements for Mixtral 8x22B Instruct V0.1 FP8?

Mixtral 8x22B Instruct V0.1 FP8 requires approximately 140.9 GB of VRAM and supports a context window of 64K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Mixtral 8x22B Instruct V0.1 FP8 and how large is it?

Mixtral 8x22B Instruct V0.1 FP8 is developed by neuralmagic, a model with 140.6b 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 Mixtral 8x22B Instruct V0.1 FP8?

Mixtral 8x22B Instruct V0.1 FP8 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Mixtral 8x22B Instruct V0.1 FP8 by neuralmagic — VRAM 140.9GB, 64K context

Name: Mixtral 8x22B Instruct V0.1 FP8
Author: neuralmagic

Mixtral 8x22B Instruct V0.1 FP8 is an open-source language model by neuralmagic. Features: 140.6b LLM, VRAM: 140.9GB, Context: 64K, License: apache-2.0, MoE, Instruction-Based, LLM Explorer Score: 0.17.

Compressed-tensors En Fp8 Instruct Mixtral Moe Region:us Safetensors Sharded Tensorflow Vllm

Model Card on HF 🤗: https://huggingface.co/neuralmagic/Mixtral-8x22B-Instruct-v0.1-FP8

Mixtral 8x22B Instruct V0.1 FP8 Parameters and Internals

Model Type

text generation

Use Cases

Areas:

commercial use, research

Applications:

assistant-like chat

Primary Use Cases:

assistant-like chat in English

Limitations:

Use in languages other than English, Use that violates applicable laws or regulations

Supported Languages

English (Proficient)

Input Output

Input Format:

Text

Accepted Modalities:

text

Output Format:

Text

Performance Tips:

Use vLLM backend for efficient deployment

Release Notes

Version:

1.1

Date:

8/11/2024

Notes:

Quantized version with FP8 optimizations

LLM Name	Mixtral 8x22B Instruct V0.1 FP8
Repository 🤗	https://huggingface.co/neuralmagic/Mixtral-8x22B-Instruct-v0.1-FP8
Model Size	140.6b
Required VRAM	140.9 GB
Updated	2025-04-08
Maintainer	neuralmagic
Model Type	mixtral
Instruction-Based	Yes
Model Files	4.9 GB: 1-of-29 4.9 GB: 2-of-29 4.9 GB: 3-of-29 4.9 GB: 4-of-29 4.9 GB: 5-of-29 4.9 GB: 6-of-29 4.9 GB: 7-of-29 4.9 GB: 8-of-29 4.9 GB: 9-of-29 4.9 GB: 10-of-29 4.9 GB: 11-of-29 4.9 GB: 12-of-29 4.9 GB: 13-of-29 4.9 GB: 14-of-29 4.9 GB: 15-of-29 4.9 GB: 16-of-29 4.9 GB: 17-of-29 4.9 GB: 18-of-29 4.9 GB: 19-of-29 4.9 GB: 20-of-29 5.0 GB: 21-of-29 5.0 GB: 22-of-29 5.0 GB: 23-of-29 4.9 GB: 24-of-29 4.9 GB: 25-of-29 4.9 GB: 26-of-29 4.9 GB: 27-of-29 4.9 GB: 28-of-29 3.4 GB: 29-of-29
Model Architecture	MixtralForCausalLM
License	apache-2.0
Context Length	65536
Model Max Length	65536
Transformers Version	4.44.0
Tokenizer Class	LlamaTokenizer
Vocabulary Size	32768
Torch Data Type	bfloat16

Rank the Mixtral 8x22B Instruct V0.1 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 Mixtral 8x22B Instruct V0.1 FP8

Best Alternatives	Context / RAM	Downloads	Likes
Dolphin 2.6 Mixtral 8x7b	32K / 93.6 GB	11285	211
Dolphin 2.6 Mixtral 8x7b	32K / 93.6 GB	7806	217
...eqlen 4096 Bs 4 Optimum 0 0 23	32K / GB	8	0
...eqlen 4096 Bs 4 Optimum 0 0 23	32K / GB	9	1
Dolphin 2.7 Mixtral 8x7b	32K / 93.6 GB	2178	170
Empower Functions Medium	32K / 93.6 GB	17	1
Mixtral 8x7B Instruct V0.1	32K / GB	5	0
...ral 8x7b Instruct V0.1 Int4 Ov	32K / 0 GB	16	4
Dolphin 2.7 Mixtral 8x7b	32K / 93.6 GB	228	169
...ct V0.1 Agent Function Calling	32K / 44.3 GB	4	2