Mixtral

Mixtral is Mistral AI’s Mixture-of-Experts model series. Each token is processed by a subset of experts, enabling a large total parameter count with efficient per-token compute.


Task	Text Generation (MoE)
Architecture	`MixtralForCausalLM`
Parameters	47B total / 13B active
HF Org	mistralai

Available Models

Mixtral-8x7B: 8 experts, 2 active per token (~13B active)
Mixtral-8x7B-Instruct: instruction-tuned variant
Mixtral-8x22B: 8 experts, 2 active per token (~39B active)

Architecture

MixtralForCausalLM

Example HF Models

Model	HF ID
Mixtral 8x7B v0.1	`mistralai/Mixtral-8x7B-v0.1`
Mixtral 8x7B Instruct v0.1	`mistralai/Mixtral-8x7B-Instruct-v0.1`

Example Recipes

Recipe	Description
mixtral-8x7b-v0-1_squad.yaml	SFT — Mixtral 8x7B on SQuAD
mixtral-8x7b-v0-1_squad_peft.yaml	LoRA — Mixtral 8x7B on SQuAD

Try with NeMo AutoModel

1. Install (full instructions):

$ pip install nemo-automodel

2. Clone the repo to get the example recipes:

$ git clone https://github.com/NVIDIA-NeMo/Automodel.git
$ cd Automodel

3. Run the recipe from inside the repo:

$ automodel --nproc-per-node=8 examples/llm_finetune/mistral/mixtral-8x7b-v0-1_squad.yaml

Run with Docker

1. Pull the container and mount a checkpoint directory:

$ docker run --gpus all -it --rm \
>   --shm-size=8g \
>   -v $(pwd)/checkpoints:/opt/Automodel/checkpoints \
>   nvcr.io/nvidia/nemo-automodel:26.06.00

2. Navigate to the AutoModel directory (where the recipes are):

$ cd /opt/Automodel

3. Run the recipe:

$ automodel --nproc-per-node=8 examples/llm_finetune/mistral/mixtral-8x7b-v0-1_squad.yaml

See the Installation Guide and LLM Fine-Tuning Guide.

Fine-Tuning

See the LLM Fine-Tuning Guide and the Large MoE Fine-Tuning Guide.