Llama 4 is not effective, Meta admits there are issues

At 1:49 am this morning, Ahmad Al Dahle, the leader of Meta Platforms (META. O) generative AI, posted a long article on social media, providing an official response to the doubts raised by Llama 4, which was just open sourced two days ago. Ahmad stated that Llama 4 was released as soon as it was developed, so there may inevitably be some differences in model quality among different services. Meta will soon fix these vulnerabilities to improve performance. At the same time, it denies conducting pre training on the test set.