Meta Launches Muse Wonder Frontier Models: Beats Opus 5.5 on SWE-Marathon, Trails by 2 Points on Intelligence Index

Meta has introduced the Muse Wonder family of frontier models, claiming state-of-the-art performance for agentic workflows in both personal and professional contexts. Benchmark data indicates that Muse Wonder outperforms Astra and Opus 5.5 on DeepSWE v1.1 and SWE-Marathon tasks, while trailing Opus 5.5 by only two points on the Artificial Analysis Intelligence Index. Despite these strong metrics, there is significant community demand for Meta to open-source the model or release a lighter version like Spark 1.3.