Meta Platforms, the owner of Facebook, is leading the open-source AI game with its Llama models.
The latest, Llama 3.1, has quickly been adopted by China’s big tech firms Alibaba and Tencent.Alibaba’s cloud platform, Alibaba Cloud, was one of the first to integrate Llama 3.1 into its Bailian model training platform.
They are offering a month of free compute resources for training and inferencing with Llama 3.1. Tencent also announced that Llama 3.1 is now available on its cloud platform.
The new model supports tasks like intelligent conversation, text generation, and writing.Meta CEO Mark Zuckerberg calls Llama 3.1 “the first frontier-level open-source AI model.”
It comes in three sizes: 8B, 70B, and 450B, with the number indicating billions of parameters.
Early tests show Llama 3.1 compares well with top closed-source models like OpenAI’s GPT-4.
Llama models are popular in China, used by many start-ups to build their AI products with custom tweaks. However, there are concerns about over-reliance on Meta’s models.
Last year, Chinese AI firm 01.AI faced criticism for using Llama models without transparency.
Zuckerberg defends open-source models, saying they benefit innovation in the US, despite the risk of espionage.
