Meta released Llama 3 in 8B and 70B parameter sizes. It is openly licensed for research and commercial use — one of the most powerful freely available AI models.
It outperforms GPT-3.5 in several benchmarks and introduced a new tokenizer with 128K vocabulary, improving efficiency for code and multilingual text.
Developers can run smaller versions locally on consumer hardware — enabling privacy-first AI applications without API fees or cloud dependency.
This significantly democratizes AI development for everyone worldwide.