David Sacks Says Meta's Llama 4 Helps the U.S. Stay Ahead in the AI Race

Web3 , Monday, 07 April 2025
Posted by Rima Dwi Astuti

Meta Launches New AI Models, Boosts U.S. in AI Race

Meta has released two new artificial intelligence (AI) models called Llama 4 Scout and Llama 4 Maverick, helping the U.S. take the lead in the global AI race. David Sacks, a top U.S. AI and crypto official, praised the move in a post on X, saying:

“For the U.S. to win the AI race, we have to win in open source too, and Llama 4 puts us back in the lead.”

Rising Competition from China’s DeepSeek

The AI race has been heating up. A Chinese startup called DeepSeek, which started in 2023, launched a powerful AI model in late 2024. Just a month later, it released a chatbot that claimed to compete with OpenAI’s ChatGPT.

DeepSeek’s AI app, DeepSeek R1, quickly became one of the most downloaded AI apps. Surprisingly, it cost only $6 million to train — much less than the $100 million reportedly spent on training ChatGPT-4. This showed that AI innovation doesn’t always need massive budgets and expensive hardware.

After DeepSeek's success, stocks of U.S. tech companies like Nvidia dropped, and some experts called this a turning point. U.S. President Donald Trump even called it a “wake-up call” for American companies.

Meta’s Llama 4 Models Explained

Meta’s new Llama 4 models are its most advanced yet. They are open-source, meaning anyone can access and use them. They also work on apps like WhatsApp and Instagram.

These models are multimodal, which means they can understand and work with different types of data — like text, images, audio, and video — all at once. This helps them better understand and respond to complex situations.

Key Features:

  • Llama 4 Scout: Has 17 billion parameters and uses 16 expert models.
  • Llama 4 Maverick: Also has 17 billion parameters but with 128 experts, making it more powerful but requiring stronger hardware.

Meta says Llama 4 Scout beats other top AI models like Gemma 3, Gemini 2.0 Flash-Lite, and Mistral 3.1. Llama 4 Maverick performs just as well as DeepSeek v3 for tasks like reasoning and coding, even though it uses fewer resources.

Meta also claims Llama 4 Maverick outperforms GPT-4o and Gemini 2.0 Flash in several benchmarks.

Looking Ahead: Llama 4 Behemoth

Meta is also working on another powerful AI model called Llama 4 Behemoth, which is still in training. The company says it could become one of the smartest AI systems in the world.

Didukung oleh
DepoCrypto.com © 2023