ModernVBERT

Model

This is the model card for modernvbert.

Table of Contents

1. Overview
2. Usage
3. Evaluation
4. License
5. Citation

Overview

The ModernVBERT suite is a suite of compact 250M-parameter vision-language encoders, achieving state-of-the-art performance in this size class, matching the performance of models up to 10x larger.

For more information about ModernVBERT, please check the arXiv preprint.

Models

• colmodernvbert (ColModernVBERT in the paper) is the late-interaction version that is fine-tuned for visual document retrieval tasks, our most performant model on this task.
• bimodernvbert (BiModernVBERT in the paper) is the bi-encoder version that is fine-tuned for visual document retrieval tasks.
• modernvbert-embed is the bi-encoder version after modality alignment (using a MLM objective) and contrastive learning, without document specialization.
• modernvbert is the base model after modality alignment (using a MLM objective).

Usage

You can use these models directly with the transformers library:

pip install torch transformers pillow

🏎️ If your GPU supports it, we recommend using ModernVBERT with Flash Attention 2 to achieve the highest GPU throughput. To do so, install Flash Attention 2 as follows, then use the model as normal:

pip install flash-attn

Here is an example of masked token prediction using ModernVBERT:

import torch
from transformers import AutoTokenizer, AutoModelForMaskedLM, AutoProcessor
from PIL import Image
from huggingface_hub import hf_hub_download

model_id = "ModernVBERT/modernvbert"

processor = AutoProcessor.from_pretrained(model_id)
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForMaskedLM.from_pretrained(
            model_id,
            torch_dtype=torch.float32, # use torch_dtype=torch.bfloat16 for flash attention
            # _attn_implementation="flash_attention_2",
            trust_remote_code=True
)

image = Image.open(hf_hub_download("HuggingFaceTB/SmolVLM", "example_images/rococo.jpg", repo_type="space"))
text = "This [MASK] is on the wall."

# Create input messages
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image"},
            {"type": "text", "text": text}
        ]
    },
]

# Prepare inputs
prompt = processor.apply_chat_template(messages)
inputs = processor(text=prompt, images=[image], return_tensors="pt")

# Inference
outputs = model(**inputs)

# To get predictions for the mask:
masked_index = inputs["input_ids"][0].tolist().index(tokenizer.mask_token_id)
predicted_token_id = outputs.logits[0, masked_index].argmax(axis=-1)
predicted_token = tokenizer.decode(predicted_token_id)
print("Predicted token:", predicted_token)
# Predicted token: painting

Evaluation

Our results can be found in the arXiv preprint. When finetuned for visual document retrieval tasks, ModernVBERT matches the performance of models nearly 10x larger on visual document benchmarks. Additionally, it provides an interesting inference speed on CPU compared to the models of similar performance.

License

We release the ModernVBERT model architectures, model weights, and training codebase under the MIT license.

Citation

If you use ModernVBERT in your work, please cite:

@misc{teiletche2025modernvbertsmallervisualdocument,
      title={ModernVBERT: Towards Smaller Visual Document Retrievers}, 
      author={Paul Teiletche and Quentin Macé and Max Conti and Antonio Loison and Gautier Viaud and Pierre Colombo and Manuel Faysse},
      year={2025},
      eprint={2510.01149},
      archivePrefix={arXiv},
      primaryClass={cs.IR},
      url={https://arxiv.org/abs/2510.01149}, 
}