Muse Wonder launches as closed-source frontier model family, breaking from Llama's open-weight tradition

A post by aiwaremax reports the launch of Muse Wonder, a new family of frontier models, with a focus on personal and professional agentic work. The post quotes an announcement from finkd: 'Introducing Muse Wonder, our new family of frontier models with state of the art performance for personal and professional agentic work.' The post also notes that this release is fully closed-source, with no weights released, marking a departure from the open-source approach of Llama models. According to the post, independent evaluation by Artificial Analysis gave Wonder Full a score of 56, matching Claude Sonnet 5.5 and exceeding GPT-6 Astra's 53. The post states API pricing is $4/$20, in the same tier as GPT-6 Sol and Claude Opus 5.5, but Wonder Full is not available via API; it is exclusive to Muse Code subscribers. The post highlights features called 'Active Memory' for cross-session long-term memory and 'Active Focus' for multi-step agent reliability, with a context window of 1 million tokens. Another post by alexandr_wang, quoting Weinbach, includes a statement that 'muse wonder seems to communicate far more naturally and than other frontier models.' alexandr_wang also writes, 'this was a big focus area for muse wonder; our goal was to build a model for everyone!' The posts do not provide an official launch date, and the reports are attributed to the respective authors.