Alexandr Wang Says Meta Is Fixing Muse Wonder Bugs 'As Fast as We Can See Them'
Hours after Monday's launch of the Muse Wonder model family, Meta's chief AI officer publicly acknowledged launch-day bugs in the model, responding to a
By Icarus2 min read
Alexandr Wang says Meta is fixing Muse Wonder’s launch-day bugs as fast as it finds them. The Meta chief AI officer acknowledged the issues publicly on X on Monday evening, posting: “we are fixing issues as fast as we can see them! there’s definitely still issues we’re trying to iron out, but thank you to everyone for giving us a bit of time to work through all the kinks 🫶”
The post came hours after Meta launched Muse Wonder and Muse Wonder Full, which Meta describes as its first frontier-class model family, on Monday. Wang was quote-posting a user who shared a before-and-after video of the same prompt run on Muse Wonder immediately after launch and again after a fix, writing that there had been “a noticeable improvement with muse-wonder.”
The acknowledgment suggests how Meta plans to operate its new frontier models: ship, take the community’s bug reports in public, and patch quickly.
A launch under a microscope
Muse Wonder arrived under unusual scrutiny. Meta claims the model is state of the art for personal and professional agent work and describes two headline capabilities: Active Memory, which carries context across sessions, and Active Focus, which keeps multi-step agents working when tasks hit obstacles. Independent benchmarker Artificial Analysis scored Muse Wonder Full at 56 on its Intelligence Index, two points behind Claude Opus 5.5 at 58; the smaller Muse Wonder scored 52.
Some of the first-day discourse centered on coding performance. When programmer Theo Browne argued that discounting Wonder for lacking “3D demo games” reinforcement learning missed the point, Wang agreed, saying Meta had aimed the model at real-world tasks and “probably could have done more” of the demo-friendly training. Shopify CEO Tobi Lutke praised the model as “very very good,” adding that Meta is “back.”
Why it matters
For a company positioning Wonder as the core of its paid Muse agent tier, launch-day bugs were bound to surface in public. Wang’s response, acknowledging them the same evening and crediting the community for the reports, signals the tone Meta wants for its frontier-model era: the fixes ship as fast as the feedback arrives.
Wang did not say which bugs were fixed, how many remain, or what caused them, leaving the scope of the fixes unanswered. Meta has said a full safety report on the model “will be available soon,” and Wang has not committed to releasing Wonder’s weights openly.
