Just now, the strongest open-source model in the US was released! In China, it would only rank seventh?

What happened: Meta released Muse Spark Open, an open-weight model with 816 billion total parameters and a 1 million token context window, aimed at coding and agent tasks. Key numbers: 816B total parameters, 420B activated per inference, 1M token context. Artificial Analysis open-weight index: 39 points, versus MiMo-V2.6-Pro 46, GLM-5.3 45, Ling-3.1-flash 41. Terminal-Bench 2.1: 76.4. Why it matters: Meta calls it its strongest US open-weight model, but third-party and benchmark scores trail several Chinese open-weight models, so its real-world reliability still needs independent testing.

Meta released Muse Spark Open, an open-weight model with 816 billion total parameters and a 1 million token context window, aimed at coding and agent tasks.

Meta calls it its strongest US open-weight model, but third-party and benchmark scores trail several Chinese open-weight models, so its real-world reliability still needs independent testing.