Llama 4

Generic Info

• Publisher: Meta AI
• Release Date: April 2025
• Models: Scout (109B/17B active), Maverick (400B/17B active)
• Context Window: 10M tokens (Scout), 1M tokens (Maverick)
• License: Llama 4 Community License
• Key Capabilities: Multimodal (Text + Images), MoE Architecture, Reasoning, Coding

Llama 4 marks Meta's leap into native multimodality with a Mixture-of-Experts architecture. Scout's 10-million token context window is unprecedented for open models, while Maverick delivers frontier performance rivaling GPT-4o and Claude Sonnet. Both models can process images alongside text.

Hello World Guide

Get started with Llama 4 using the Hugging Face transformers library.

import torch
from transformers import pipeline

model_id = "meta-llama/Llama-4-Scout-17B-16E-Instruct"

pipe = pipeline(
    "text-generation",
    model=model_id,
    model_kwargs={"torch_dtype": torch.bfloat16},
    device_map="auto",
)

messages = [
    {"role": "system", "content": "You are a helpful assistant."},
    {"role": "user", "content": "Explain quantum entanglement in simple terms."},
]

outputs = pipe(
    messages,
    max_new_tokens=256,
)

print(outputs[0]["generated_text"][-1])

Industry Usage