Meta: Llama 4 Maverick

Overview

Open-weight Llama 4 MoE model (17B active, 400B total, 128 experts) with text and image input.

Provider

MEMeta

Modality

Text and image

Input

text, image

Output

text

Weights

Open weights

Supported parameters

toolsProvider's model page, checked October 8, 2026.

Distillable

Conditional (licence terms)Llama 4 Community License: a model trained on its outputs that is distributed must start its name with "Llama" and show "Built with Llama". The licence's other terms apply.Licence or terms, checked October 8, 2026.

Tags

open-weights, multimodal, multilingual

Docs

Provider documentation

Data policy

Prompts go to

Wherever you or your chosen host run the weights (no Meta-hosted API for this model)

Retention

Set by the host, Meta does not receive prompts when you self-host

Zero data retention

Not stated by the provider

Region

Depends on host

Policy

Meta data policy

Summarised from the provider's published terms. The provider's own policy is what applies.

Where you can run it

How Llama 4 Maverick can be deployed, bought and reached through Safeguard.

Deployment

Private cloudOn-premAir-gapped

Procurement

Not published yet

Inference regions

Not published yet

On-prem and air-gapped deployments run wherever you install them.

Pricing

Notes

Price on request: Meta publishes no first-party USD per-token price for Llama models.

Source

Meta pricing

No markup on model usage: you pay the provider's list price.

API

OpenAI-compatible. Your API key and access to this model are set up for your workspace when you choose Use with Safeguard.

from openai import OpenAI

client = OpenAI(
    base_url="https://api.safeguard.sh/gpt/v1",
    api_key="SAFEGUARD_API_KEY",
)

response = client.chat.completions.create(
    model="meta/llama-4-maverick",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Base URL https://api.safeguard.sh/gpt/v1, model meta/llama-4-maverick.