Meta's open-weights multimodal LLM
Category: Text & Copy. Pricing: free. Last modified: 2026-09-07.
Meta's flagship open-weights LLM family, Scout (17B active) and Maverick, the first natively multimodal Llama models with mixture-of-experts architecture and unprecedented context length.
Not editorially tested: One AI Guide has not published a complete hands-on test record for this tool.
Published by LUMIVEX for One AI Guide.
Editorial lead: Said El Moussaoui. Our methodology is published at /how-we-rate.
Official vendor website: see the vendor link below.
Llama 4 is a text & copy tool. Meta's open-weights multimodal LLM.
The One AI Guide catalog currently records Llama 4 as free. This is a directory summary, not a current vendor quote; always check the vendor's site for the latest plans before signing up.
The catalog currently records Llama 4's pricing model as Free. Check the vendor's site because availability, limits, and plan terms can change.
Common use cases include Self-hosted deployments where data must not leave your infra, and Fine-tuning a base model for a specific domain.
The One AI Guide catalog currently highlights these strengths: Open weights up to 405B parameters, full commercial use; Strong base for fine-tuning, thousands of community variants; Runs on consumer hardware (8B at 16GB VRAM).