Meta Releases Llama 3.1 to Challenge Closed AI Models
Meta unveils Llama 3.1, a massive 405-billion-parameter open-source AI model that directly rivals top closed systems from OpenAI and Anthropic. The updated model family features a massive 128,000-token context window and broad multilingual support.
Meta unveils Llama 3.1, its largest open-source artificial intelligence model to date, featuring a massive 405 billion parameters. The company claims this flagship model rivals the most powerful closed systems on the market, including GPT-4o and Claude 3.5 Sonnet, across tasks like general knowledge, math, and tool use.
The new model boasts an expanded 128,000-token context window that allows it to process the equivalent of a 400-page novel in a single prompt. Meta also adds support for eight new languages, including French, German, Hindi, and Spanish, bringing advanced multilingual translation and long-form summarization to all model sizes.
Alongside the massive 405B release, Meta updates its 8 billion and 70 billion parameter models with the same context window and language capabilities. The company updates its licensing to allow developers to use Llama 3.1 outputs to train and fine-tune smaller, more specialized AI models.