Meta releases the new Llama 3.1 models family, and they surpass every possible AI tests

Meta has taken the AI world by storm with its new invention, the Llama 3.1 AI models family. It includes Llama 3.1 8B, Llama 3.1 70B, and the massive Llama 3.1 405B, all with an enhanced context length of up to 128 K each.

They are available for use, and according to the announcement, developers can utilize the outputs of these Llama models to improve other models, expanding the limits of what AI can do. Meta subjected Llama 3.1 to more than 150 benchmark datasets and even human evaluations to measure its performance against other top models in real-life situations.

So, what is the result? The Llama 3.1 405B model, like its older sibling GPT-4 (which is already old, as OpenAI is seeking to release the next model), is a giant in open-source models and not only the biggest but also one of the most powerful foundation models for everyone to use. Even lesser versions of the Llama 3.1 family can fight equally with both closed and open models that are similar in size.

The Llama 3.1 family can be reached through AWS, NVIDIA, Databricks, and other important platforms. Additionally, Azure AI has presented the 405B model as a serverless API endpoint. Thanks to Meta’s dedication to open-source AI, the new Llama family is more accessible than ever.

Even Mark Zuckerberg mentioned how quickly open-source models are advancing—not only matching but often surpassing closed models in terms of being open, modifiable, and cost-effective.

Today, several tech companies are developing leading closed models. But open source is quickly closing the gap. Last year, Llama 2 was only comparable to an older generation of models behind the frontier. This year, Llama 3 is competitive with the most advanced models and leading in some areas. Starting next year, we expect future Llama models to become the most advanced in the industry. But even before that, Llama is already leading on openness, modifiability, and cost efficiency.

Mark Zuckerberg

The company wants to make a stronger AI model, spread AI to many people, encourage creative ideas, and establish fresh norms for cooperation and clarity in the area, and it might just do that with Llama 3.1. So, what is your opinion? Are we witnessing a pivotal moment in the evolution of AI models?

Check out the full blog post if you want an in-depth analysis of the latest Llama family.