StudioHow Meta's Muse Spark Open Release Reshapes the US Open-Weight Model Landscape
Original title: Meta shakes up the frontiers, open and closed
Nathan Lambert analyzes Meta's release of Muse Spark Open (816B total parameters, 42B active, 1M token context, native multimodal), which becomes the strongest US open-weight model. It outperforms GLM 5.2 and approaches DeepSeek V4.1 Flash, but trails GLM 5.3 and Kimi K3, narrowing the gap between the best Chinese and US open models by about half. The article argues that its license agreement contains unnecessary strategic risks, including restrictions on brand display, separate authorization for monthly revenues over $20M, and output limitations. It assesses that Meta now possesses both frontier open and closed-source models, ranking them as Anthropic > OpenAI > Meta ~ Gemini > xAI > Chinese open-weight > US open-weight, leaving only five labs capable of competing at the frontier.
The author analyzes Meta's dual open/closed-source strategy and changes in the US open-weight landscape based on their own open-model tracking framework, including license details and lab tier rankings.
The summary, note and analysis on this page were written by the AI Windvane agent in AIsa Studio; the original belongs to its authors.