Hasty Briefsbeta

Bilingual

Meta's surprise Llama 4 drop exposes the gap between AI ambition and reality

a year ago
  • #Meta
  • #AI
  • #Multimodal
  • Meta released Llama 4 multimodal AI models, including Scout and Maverick, with a 10 million token context window for Scout.
  • Initial reception from the AI community has been mixed-to-negative, with criticism of Meta's marketing versus user experience.
  • Meta describes Llama 4 as 'natively multimodal,' using 'early fusion' to handle text and images, competing with OpenAI's GPT-4o and Google's Gemini 2.5.
  • The models were trained with assistance from an unreleased 'teacher' model, Llama 4 Behemoth, which has 2 trillion parameters and is still in development.
  • Meta continues to use the term 'open source' despite licensing restrictions, with 'open weights' being a more accurate description.