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

A post from the account aiwaremax reports the launch of Muse Wonder, a new family of frontier models, and highlights a decisive shift: the models are fully closed-source with no weights released, marking a clear departure from the open-source approach associated with Llama. The post cites an independent evaluation by Artificial Analysis that gave Wonder Full a score of 56, tying Claude Sonnet 5.5 and exceeding GPT-6 Astra's 53. API pricing is listed at $4/$20, in line with GPT-6 Sol and Claude Opus 5.5, but Wonder Full is not available via API; it is offered only to Muse Code subscribers. The post also mentions 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. A quoted announcement from finkd introduces Muse Wonder as a new family of frontier models with state-of-the-art performance for personal and professional agentic work. Another post from alexandr_wang, quoting Weinbach, states that communication was a big focus area for Muse Wonder, with the goal of building a model for everyone, and includes a quoted remark that Muse Wonder seems to communicate far more naturally than other frontier models. These reports are based on the authors' claims and have not been independently verified.