Meta's Muse Wonder model promoted as fast and token-efficient

Alexandr Wang promotes Meta's Muse Wonder model on X as "super fast and token efficient" and invites readers to "give it a go." His post gives no benchmark figures.

A separate post by @merge_api, citing Artificial Analysis data for the week of Oct 12, reports that Muse Wonder averages 6.8 minutes per task with a 52.4 Intelligence Index score. The same post reports 13.2 minutes and a 57.6 score for Claude Opus 5.5, and 15.3 minutes and a 56.0 score for Sonnet 5.5. Wang's claim of speed and efficiency is his own; the timing figures come from the @merge_api post.