Meta introduces Muse Wonder AI suite, taking aim at industry leaders
The social media giant has rolled out Muse Wonder and Wonder Full, bringing enhanced memory and reasoning capabilities as the flagship offering surpassed Astra and Fable in AI benchmark evaluations.
The tech giant has introduced a pair of cutting-edge artificial intelligence offerings, Muse Wonder and Muse Wonder Full, describing them as the company's most advanced models yet and positioning them as progress toward achieving "personal superintelligence."
The company announced on Monday that these models feature two new functionalities dubbed Active Memory and Active Focus by Meta. The former maintains pertinent details throughout ongoing dialogues, while the latter enables the system to maintain focus when tackling complex challenges.
The independent assessment platform Artificial Analysis gave Wonder Full a score of 56 on its Intelligence Index, while the base Wonder version received 52, with both evaluations conducted at maximum reasoning effort.
This positions Full ahead of Claude Fable 5.1 and GPT-6 Astra, both of which achieved scores of 53, though it remains behind Claude Opus 5.5's score of 58.
Bindu Reddy, CEO of Abacus.AI, which performs independent model assessments, disputed the performance assertions, stating on X that Muse Wonder was "nowhere near Fable or Astra level" and indicated that comprehensive benchmark findings would be released subsequently.
The company's latest models represent its priciest offerings yet, with the base Wonder model priced at $4 per million input tokens, equivalent to Claude Opus 5.5 pricing but offering better value than GPT-6 Astra and Claude Fable 5.1, which are priced at $10 and $50 respectively.