What are the hardware requirements for Smaug Mixtral V0.1?

Smaug Mixtral V0.1 requires approximately 187.7 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 Smaug Mixtral V0.1 and how large is it?

Smaug Mixtral V0.1 is developed by abacusai, a model with 46.7b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Smaug Mixtral V0.1 perform on standard benchmarks?

Smaug Mixtral V0.1 has the following published scores: MMLU 70.06. Compare against reference models on this page or on the LLM Explorer leaderboards.

Smaug Mixtral V0.1 by abacusai — VRAM 187.7GB, 32K context

Name: Smaug Mixtral V0.1
Author: abacusai

Smaug Mixtral V0.1 is an open-source language model by abacusai. Features: 46.7b LLM, VRAM: 187.7GB, Context: 32K, License: apache-2.0, Fine-Tuned, LLM Explorer Score: 0.21, Arc: 74.7, HellaSwag: 87.7, MMLU: 70.1, GSM8K: 72.

Arxiv:2402.13228 Conversational Endpoints compatible Finetuned Mixtral Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/abacusai/Smaug-Mixtral-v0.1

Smaug Mixtral V0.1 Parameters and Internals

Model Type

finetune

Additional Notes

This model is designed for low edit distance datasets, like math-based datasets, and performs well across diverse tasks due to the DPOP training technique.

Training Details

Data Sources:

ARC, HellaSwag, MetaMath, other existing datasets

Methodology:

DPO-Positive (DPOP), pairwise preference versions

LLM Name	Smaug Mixtral V0.1
Repository 🤗	https://huggingface.co/abacusai/Smaug-Mixtral-v0.1
Model Size	46.7b
Required VRAM	187.7 GB
Updated	2026-07-24
Maintainer	abacusai
Model Type	mixtral
Model Files	4.9 GB: 1-of-39 4.9 GB: 2-of-39 4.9 GB: 3-of-39 4.9 GB: 4-of-39 4.9 GB: 5-of-39 4.9 GB: 6-of-39 4.9 GB: 7-of-39 4.9 GB: 8-of-39 4.9 GB: 9-of-39 4.9 GB: 10-of-39 4.9 GB: 11-of-39 5.0 GB: 12-of-39 4.8 GB: 13-of-39 4.9 GB: 14-of-39 4.9 GB: 15-of-39 4.9 GB: 16-of-39 4.9 GB: 17-of-39 4.9 GB: 18-of-39 4.9 GB: 19-of-39 4.9 GB: 20-of-39 4.9 GB: 21-of-39 4.9 GB: 22-of-39 4.9 GB: 23-of-39 4.9 GB: 24-of-39 4.9 GB: 25-of-39 4.9 GB: 26-of-39 4.9 GB: 27-of-39 4.9 GB: 28-of-39 4.9 GB: 29-of-39 4.9 GB: 30-of-39 4.9 GB: 31-of-39 4.9 GB: 32-of-39 4.9 GB: 33-of-39 4.9 GB: 34-of-39 4.9 GB: 35-of-39 4.9 GB: 36-of-39 5.0 GB: 37-of-39 4.8 GB: 38-of-39 1.5 GB: 39-of-39
Model Architecture	MixtralForCausalLM
License	apache-2.0
Context Length	32768
Model Max Length	32768
Transformers Version	4.36.2
Tokenizer Class	LlamaTokenizer
Vocabulary Size	32000
Torch Data Type	float32

Rank the Smaug Mixtral V0.1 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 Smaug Mixtral V0.1

Best Alternatives	Context / RAM	Downloads	Likes
Mixtral 8x7B Instruct V0.1	32K / 93.6 GB	842893	4720
Nous Hermes 2 Mixtral 8x7B DPO	32K / 93.6 GB	40865	457
Mixtral 8x7B V0.1	32K / 93.6 GB	49231	1820
Sensualize Mixtral Bf16	32K / 93.6 GB	0	0
Skadi Mixtral V1	32K / 93.5 GB	0	0
Franziska Mixtral V1	32K / 93.5 GB	0	0
Typhon Mixtral V1	32K / 93.4 GB	0	0
GritLM 8x7B KTO	32K / 93.6 GB	7976	3
Mixtral 8x7B Instruct V0.1	32K / 93.6 GB	7	0
Mixtral 8x7B Instruct V0.1 FP8	32K / 47.1 GB	3383	0

Note: green Score (e.g. "73.2") means that the model is better than abacusai/Smaug-Mixtral-v0.1.