Phi-4-multimodal | NVIDIA NeMo AutoModel

Phi-4-multimodal-instruct is Microsoft’s multimodal extension of Phi-4, supporting text, image, and audio inputs — making it suitable for speech, vision, and combined multimodal tasks.


Task	Omnimodal (Text·Image·Audio)
Architecture	`Phi4MultimodalForCausalLM`
Parameters	5.6B
HF Org	microsoft

Available Models

Phi-4-multimodal-instruct: 5.6B

Architecture

Phi4MultimodalForCausalLM

Example HF Models

Model	HF ID
Phi-4-multimodal-instruct	`microsoft/Phi-4-multimodal-instruct`

Example Recipes

Recipe	Dataset	Description
phi4_mm_cv17.yaml	CommonVoice 17	SFT — Phi-4-multimodal on CommonVoice (audio-text)

Try with NeMo AutoModel

1. Clone and install from source (full instructions):

git clone https://github.com/NVIDIA-NeMo/Automodel.git
cd Automodel
uv sync --locked --all-groups --all-extras

2. Run the recipe from inside the repo:

uv run automodel --nproc-per-node=8 examples/vlm_finetune/phi4/phi4_mm_cv17.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/vlm_finetune/phi4/phi4_mm_cv17.yaml

See the Installation Guide and Omni Fine-Tuning Guide.

Fine-Tuning

See the VLM / Omni Fine-Tuning Guide.

Hugging Face Model Cards

microsoft/Phi-4-multimodal-instruct