Meta Launches Muse Wonder AI Models to Challenge OpenAI and Anthropic

Meta has rolled out two fresh artificial intelligence models named Muse Wonder and Muse Wonder Full, describing them as its strongest creations so far. Company founder Mark Zuckerberg framed the release as progress toward what he calls “personal superintelligence.”

Active Memory and Active Focus Features

The models bring two new functions Meta has labeled Active Memory and Active Focus. Active Memory keeps important details alive across separate chats, while Active Focus allows the system to keep pushing through tough, long-running jobs.

Benchmark Scores and Pricing Details

Independent testing group Artificial Analysis gave Wonder Full a 56 on its Intelligence Index, with standard Wonder at 52. That places Full ahead of Claude Fable 5.1 and GPT-6 Astra, both at 53, though still behind Claude Opus 5.5 at 58.

Not everyone is convinced. Bindu Reddy, head of Abacus.AI, pushed back on X, claiming Muse Wonder is “nowhere near Fable or Astra level” and promising fuller benchmark data soon.

Pricing is steep. Standard Wonder costs $4 per million input tokens, equal to Claude Opus 5.5 and cheaper than GPT-6 Astra and Claude Fable 5.1. Wonder Full, however, is locked behind Muse Code subscriptions of $50 or $200 monthly.