Llama 3.1, this giant open source language model with 405 billion parameters, caused a huge shock in the AI field due to leaks without official release. Its performance is so powerful that it even surpasses GPT-4o in some benchmark tests, setting a new benchmark for open source models. The heated discussion on Reddit further proves its impact on the AI community. This article will delve into the performance, highlights, and safety measures of Llama 3.1 and unveil this mysterious model.
Llama3.1 has been leaked! You heard it right, this open source model with 405 billion parameters has caused an uproar on Reddit. This is probably the closest open source model to GPT-4o to date, and even surpasses it in some aspects.
Llama3.1 is a large language model developed by Meta (formerly Facebook). Although there is no official release yet, the leaked version has already caused a stir in the community. This model includes not only the base model, but also benchmark results of 8B, 70B and the maximum parameter of 405B.
Performance comparison: Llama3.1 vs GPT-4o
Judging from the leaked comparison results, even the 70B version of Llama3.1 surpassed GPT-4o in multiple benchmark tests. This is the first time that an open source model has reached the SOTA (State of the Art, the most advanced technology) level on multiple benchmarks. People can’t help but sigh: The power of open source is really powerful!
Model highlights: multi-language support, richer training data
The Llama3.1 model uses 15T+ tokens from public sources for training, and the pre-training data deadline is December 2023. It supports not only English but also French, German, Hindi, Italian, Portuguese, Spanish and Thai. This makes it great in multilingual conversation use cases.
The Llama3.1 research team attaches great importance to the security of the model. They used a multifaceted data collection approach that combined human-generated and synthetic data to mitigate potential security risks. In addition, the model also introduces boundary prompts and adversarial prompts to enhance data quality control.
Model card source: https://pastebin.com/9jGkYbXY#google_vignette
The leak of Llama 3.1 will undoubtedly have a profound impact on the AI field. It not only demonstrates the huge potential of open source models, but also triggers further thinking about model security and ethical issues. In the future, we will continue to pay attention to Llama 3.1 and its subsequent development, and look forward to it bringing more surprises to the advancement of AI technology.