Skip to content

Pick your style: Denkbaar

Book a conversation

Articles

History

Llama 4 arrives with Scout and Maverick

·

Meta introduced two models that process text and images using a mixture of experts.

A developer looks at a screen with an image and a document beside computing modules.
Llama 4 combined image processing with available model weights. Illustration: Denkbaar.

On 5 April 2025, Meta introduced Llama 4 Scout and Maverick. Both models used a mixture-of-experts design: several components, with a subset active for each input. Meta described the models as handling text and images. The announcement contains the design and comparisons. A screen with a plant and a document is connected to a few coloured computing modules.

We see the main historical point as the combination of available model weights and image processing. This gave builders another reason to investigate custom applications involving documents or images. Meta’s comparisons remain claims from the maker. Two hands compare a practice image with a sheet and a laptop screen.

For a small or medium-sized business, one possible trial is describing public practice images. Supply fixed questions: what is visible, where is it and what cannot be established? Prepare a list of correct answers beforehand. Ask the builder to explain which model configuration and equipment are needed. Check the usage terms for the chosen application separately. Then assess missing and invented details in the answers. The useful outcome is an appropriately tested component for your work, with a clear view of maintenance and the limitations of that particular installation.

Sources

  1. Llama 4 Scout and Maverick

Topics: open source