AI Daily Digest October 17, 2026: Meta Opens 816B-Parameter Model

Meta released Muse Spark Open, an open-weights sparse MoE model with 816B total parameters activating ~42B per token, a 1M-token context and text, image, video and audio input. NVIDIA shipped Nemotron 3.5 Super VL at 120B parameters, and ChatGPT can now read and act inside open Chrome tabs.

• Muse Spark Open: 816B parameters, ~42B active per token, 1M-token context
• Runs locally on two Mac Studio machines with M5 Ultra and 512GB memory each
• Nemotron 3.5 Super VL: 120B parameters, 12B active, OpenMDW-1.1 license
• ChatGPT reads open Chrome tabs and acts in them via an extension