Meta defends Llama 4 release against ‘reports of mixed quality,’ blames bugs

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Meta’s new flagship AI language model Llama 4 came suddenly over the weekend, with the parent company of Facebook, Instagram, WhatsApp and Quest VR (among other services and products) revealing not one, not two, but three versions — all upgraded to be more powerful and performant using the popular “Mixture-of-Experts” architecture and a new training method involving fixed hyperparameters, known as MetaP. Also, all three are equipped with massive context windows — the amount of information that an AI language model can handle in one input/output exchange with a user or tool. But following the surprise announcement and public release of two of those models for download and usage — the lower-parameter Llama 4 Scout and mid-tier Llama 4 Maverick — on Saturday, the response from the AI community on social media has been less than adoring. Llama 4 sparks confusio ... Read full article.