Meta has released Muse Wonder, described as its first frontier model. Unlike Muse Spark, which was positioned as a lightweight model focused on low latency, the Wonder models are larger and said to be more capable. Meta reports higher benchmark scores and improved performance on complex tasks such as coding.
Why it matters
The release marks Meta’s entry into the frontier model tier, shifting from a lightweight, latency-oriented offering toward larger models aimed at more demanding workloads.
Who should care
Developers and teams working on complex tasks like coding may be interested in evaluating the Wonder models against the lighter Muse Spark.