Meta takes another step forward in artificial intelligence. The company has unveiled Muse Wonder and Muse Wonder Full, its first two frontier models and an evolution of Muse Spark 1.3 designed to compete with the best AI on the market. As reported by TechCrunch, Meta says both models are on par with Fable 5.1 and GPT-6 Astra in several internal tests, and the results are genuinely impressive.
Muse Wonder takes direct aim at GPT-6 Astra
Muse Wonder is designed from the outset for complex tasks, especially for coding and AI agent use cases. Compared with the much faster Muse Spark approach, Meta is prioritizing greater capability at a larger scale. Muse Wonder will cost $4 per million input tokens and $20 per million output tokens through the Meta Model API.
Wonder Full, meanwhile, will be available to Muse Code subscribers. Both models will, of course, power the Muse app, which has already become one of the most-downloaded AI apps.
If there’s one standout among Meta’s new additions, it’s Active Memory. Meta wants Muse Wonder to remember relevant information for days, weeks, or months. On MemoryBench, Muse Wonder outperforms Fable, Opus, and Astra according to Meta’s tests.
Today’s launches put Meta at the forefront. GPT-6 Astra is currently the performance benchmark, while Apple retains a different kind of edge thanks to the integration of Apple Intelligence with its devices and its security and privacy mechanisms, but rivals are starting to deliver increasingly capable, better-integrated tools.
