The successor to Meta’s open source model Llama 2 is finally here: Llama 3. Meta AI is also seeing the light of day, but not in Europe yet.

Meta launches Llama 3, the successor to its open source model Llama 2. Llama 3 consists of two models, Llama 3-8B and Llama 3-70B. In addition, Meta is also launching Meta AI, an AI assistant that the company says is one of the best in the world.

Two models

The launch of Llama 3 consists of two models: Llama 3-8B (which contains 8 billion parameters) and Llama 3-70B (which contains 70 billion parameters). The models are trained on two custom-built 24,000 GPU clusters, making them the most powerful generative AI models currently available, according to the company.

The AI ​​benchmarks show how the two new models perform alongside the existing models from Google, OpenAI, Anthropic and Mistral. Both variants seem to perform well, with the 70B variant producing even more impressive results. Compared to its predecessor Llama 2, which can process up to 32,000 tokens, Llame 3 performs better with 128,000 tokens. Developers can customize the model as desired using tools.

Meta AI

Simultaneously with the Llama 3 model, Meta immediately introduces a new application that can run on the new model: Meta AI. This AI-powered chatbot can be used on Facebook, Instagram, WhatsApp and Messenger. Meta AI works multimodally and can generate images in addition to text prompts. The chatbot is currently not yet available in Europe.