Meta Llama 4

vLlama 4 (Scout, Maverick)FreeReleased 5 Apr 2025

Meta's open-weight Llama 4 models, Scout and Maverick, for self-hosting. By Meta Platforms, Inc.

Web app

Official sourceNo bundles

By the ToolsBharat teamUpdated 5 Apr 2025How we check

Quick answer

Meta Llama 4 is a free web app from Meta Platforms, Inc.. Meta's open-weight Llama 4 models, Scout and Maverick, for self-hosting. Download it only from huggingface.co. The latest version we track is Llama 4 (Scout, Maverick) (5 Apr 2025). ToolsBharat links straight to the publisher's own download page, with no bundled offers.

Latest version:: Llama 4 (Scout, Maverick) (5 Apr 2025)
Publisher:: Meta Platforms, Inc.
Official site:: dev.meta.ai
Licence:: Free — Llama 4 Community License Agreement (custom; free commercial use, separate licence needed above 700 million monthly active users)
Requirements:: Meta says Scout fits on a single NVIDIA H100 GPU with on-the-fly int4 quantisation, and Maverick's FP8 weights fit on a single H100 DGX host

Our verdict

Llama 4 brings mixture-of-experts designs with very long context windows, and Hindi is one of its twelve officially supported languages. Even the smaller Scout model has 109B total parameters, so it is a workstation or server model rather than something for an ordinary laptop. Downloads sit behind a licence form on Hugging Face, and the licence is Meta's own, not an open-source one.

Key features

Llama 4 Scout: 17B active of 109B total parameters, 16 experts, 10M-token context
Llama 4 Maverick: 17B active of 400B total parameters, 128 experts, 1M-token context
Text and image input
Twelve officially supported languages, including Hindi
BF16 weights, plus FP8 weights for Maverick

Pros

Pro: Hindi is an officially supported language
Pro: Very long context windows
Pro: Commercial use allowed for most businesses

Cons

Con: Gated: you must accept Meta's licence and submit a form with your legal name, date of birth and organisation
Con: Custom licence with an acceptable-use policy, not OSI open source
Con: Too large for typical consumer PCs; Meta's own hardware notes refer to H100-class GPUs

What's new in Llama 4 (Scout, Maverick)

See what changed in Llama 4 (Scout, Maverick) in the official release notes

Version history

Version Llama 4 (Scout, Maverick)Latest
5 Apr 2025
SHA-256
The developer does not publish one
Source
Release notes

Requirements

OS: Meta says Scout fits on a single NVIDIA H100 GPU with on-the-fly int4 quantisation, and Maverick's FP8 weights fit on a single H100 DGX host
Architecture: Model weights (runs via a runtime such as Ollama, LM Studio or llama.cpp)
Also needs: Free Hugging Face account and accepted licence terms required to download

Licence

Type: Free
Licence: Llama 4 Community License Agreement (custom; free commercial use, separate licence needed above 700 million monthly active users)
Publisher: dev.meta.ai

Where these facts come from

Checked 30 Oct 2026

FAQ

How do I run Llama 4 on my own computer?

The easiest route is Ollama: the Ollama library lists llama4, with the Scout build at about 67 GB and Maverick at about 245 GB. LM Studio and llama.cpp can run quantised GGUF builds too. Either way you need a machine with a lot of memory.

Is Llama open source?

Not in the OSI sense. The weights are free to download and use commercially, but under Meta's Llama 4 Community License, which has usage rules and a separate licence requirement for very large companies.

User reviews

User reviewsयूज़र रिव्यू

No reviews yet

Used Meta Llama 4? Reviews appear after moderation; we never import or invent ratings.

Alternatives