Meta Llama Llama 4 Maverick 17B 128E Instruct

🚀 Llama 4 Models

The Llama 4 models are natively multimodal AI models, enabling text and multimodal experiences. They use a mixture-of-experts architecture to achieve industry-leading performance in text and image understanding.

🚀 Quick Start

Prerequisites

Please ensure you have transformers v4.51.0 installed. You can upgrade it using the following command:

pip install -U transformers

Usage Example

from transformers import AutoProcessor, Llama4ForConditionalGeneration
import torch

model_id = "meta-llama/Llama-4-Maverick-17B-128E-Instruct"

processor = AutoProcessor.from_pretrained(model_id)
model = Llama4ForConditionalGeneration.from_pretrained(
    model_id,
    attn_implementation="flex_attention",
    device_map="auto",
    torch_dtype=torch.bfloat16,
)

url1 = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/0052a70beed5bf71b92610a43a52df6d286cd5f3/diffusers/rabbit.jpg"
url2 = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/datasets/cat_style_layout.png"
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": url1},
            {"type": "image", "url": url2},
            {"type": "text", "text": "Can you describe how these two images are similar, and how they differ?"},
        ]
    },
]

inputs = processor.apply_chat_template(
    messages,
    add_generation_prompt=True,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
).to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=256,
)

response = processor.batch_decode(outputs[:, inputs["input_ids"].shape[-1]:])[0]
print(response)
print(outputs[0])

✨ Features

• Native Multimodality: The Llama 4 models support text and multimodal experiences, offering industry-leading performance in text and image understanding.
• Mixture-of-Experts Architecture: They leverage a mixture-of-experts architecture to provide high performance.
• Two Efficient Models: The Llama 4 series includes Llama 4 Scout (17 billion parameters, 16 experts) and Llama 4 Maverick (17 billion parameters, 128 experts).

📦 Installation

To use the Llama 4 models with the transformers library, make sure you have transformers v4.51.0 installed. You can upgrade it using the following command:

pip install -U transformers

💻 Usage Examples

Basic Usage

from transformers import AutoProcessor, Llama4ForConditionalGeneration
import torch

model_id = "meta-llama/Llama-4-Maverick-17B-128E-Instruct"

processor = AutoProcessor.from_pretrained(model_id)
model = Llama4ForConditionalGeneration.from_pretrained(
    model_id,
    attn_implementation="flex_attention",
    device_map="auto",
    torch_dtype=torch.bfloat16,
)

url1 = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/0052a70beed5bf71b92610a43a52df6d286cd5f3/diffusers/rabbit.jpg"
url2 = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/datasets/cat_style_layout.png"
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": url1},
            {"type": "image", "url": url2},
            {"type": "text", "text": "Can you describe how these two images are similar, and how they differ?"},
        ]
    },
]

inputs = processor.apply_chat_template(
    messages,
    add_generation_prompt=True,
    tokenize=True,
    return_dict=True,
    return_tensors="pt",
).to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=256,
)

response = processor.batch_decode(outputs[:, inputs["input_ids"].shape[-1]:])[0]
print(response)
print(outputs[0])

📚 Documentation

Model Information

Intended Use

• Intended Use Cases: Commercial and research use in multiple languages. Instruction tuned models are for assistant-like chat and visual reasoning tasks, while pretrained models can be adapted for natural language generation. For vision, the models are optimized for visual recognition, image reasoning, captioning, and answering general questions about an image. The Llama 4 Community License allows for these use cases.
• Out-of-scope: Use that violates applicable laws or regulations, the Acceptable Use Policy, or the Llama 4 Community License. Use in languages or capabilities beyond those explicitly supported.

Hardware and Software

• Training Factors: Custom training libraries, Meta's custom built GPU clusters, and production infrastructure were used for pretraining. Fine-tuning, quantization, annotation, and evaluation were also performed on production infrastructure.
• Training Energy Use: Model pre-training utilized a cumulative of 7.38M GPU hours of computation on H100-80GB (TDP of 700W) type hardware.
• Training Greenhouse Gas Emissions: Estimated total location-based greenhouse gas emissions were 1,999 tons CO2eq for training. Market-based emissions were 0 tons CO2eq.

Benchmarks

Pre-trained models

Instruction tuned models

^reported numbers for MMMU Pro is the average of Standard and Vision tasks

Quantization

The Llama 4 Scout model is released as BF16 weights and can fit within a single H100 GPU with on-the-fly int4 quantization. The Llama 4 Maverick model is released as both BF16 and FP8 quantized weights. The FP8 quantized weights fit on a single H100 DGX host while maintaining quality.

🔧 Technical Details

Training Methodology

The Llama 4 models were pretrained using custom training libraries, Meta's custom built GPU clusters, and production infrastructure. Fine-tuning, quantization, annotation, and evaluation were also performed on production infrastructure.

Energy Use and Emissions

Model pre-training utilized a cumulative of 7.38M GPU hours of computation on H100-80GB (TDP of 700W) type hardware. Estimated total location-based greenhouse gas emissions were 1,999 tons CO2eq for training, while market-based emissions were 0 tons CO2eq.

📄 License

The Llama 4 models are licensed under the Llama 4 Community License Agreement. View license

⚠️ Important Note

The information you provide will be collected, stored, processed and shared in accordance with the Meta Privacy Policy.

💡 Usage Tip

1. Llama 4 has been trained on a broader collection of languages than the 12 supported languages. Developers may fine-tune the models for additional languages, but they must comply with the Llama 4 Community License and the Acceptable Use Policy.
2. Llama 4 has been tested for image understanding up to 5 input images. If using additional image understanding capabilities, developers should ensure their deployments are mitigated for risks and perform additional testing and tuning.