Meta’s open-weight AI model handles text, images, video, sound, coding, and agent work.
It is a help desk with 800 billion workers. For each question, only 40 billion wake up. The rest are on coffee break.
It can read text, pictures, video, and sound. It helps with long files, coding, and tool jobs with many steps.
MoE
It uses MoE, so only some experts wake up for each token.
Multimodal AI
It can take text, images, video, and audio as input.
Agentic coding
It handles coding jobs with many steps. It can use tools.
Open weights
Its weights can be downloaded and run on your own setup.