Llama 4 Scout and Maverick Fuel US Edge in AI Competition

On April 5, he posted on X that Meta’s new model, Llama 4, has helped the country regain its lead—especially in open-source development. He wrote, "For the US to win the AI race, we have to win in open source too, and Llama 4 puts us back in the lead".

Sacks' latest comments followed Meta’s announcement of two new AI models: Llama 4 Scout and Llama 4 Maverick.

Did you know?

Want to get smarter & wealthier with crypto?

Subscribe - We publish new crypto explainer videos every week!

Best Crypto Analysis Indicators Explained (Beginner-Friendly Animation)

SUBSCRIBE

ON YOUTUBE

Both were introduced in an April 5 post by Meta’s AI team, which described them as the company’s most capable models so far, designed to work well across different types of tasks.

According to Meta, the Scout model uses 17 billion active parameters and works with 16 experts. It is said to perform better than other models such as Gemma 3, Gemini 2.0 Flash-lite, and Mistral 3.1 on several common benchmarks used to compare AI tools.

The Maverick version also uses 17 billion active parameters but is built with 128 experts instead. Meta claims this version beats GPT-4o and Gemini 2.0 Flash in similar benchmark tests. It also says Maverick can match DeepSeek v3 when it comes to tasks like logical reasoning and code generation, even though it uses fewer active parameters than DeepSeek.

Trump administration officials, including David Sacks, recently met with senior figures from the United Arab Emirates (UAE). What did they talk about? Read the full story.