Meta says it is expanding its AI model lineup with new releases targeting coding and agentic tasks. Across reporting, Meta introduces models branded around Muse Spark Open and Muse Wonder, describing them as capable of handling complex work such as software development, document analysis, and research.
One outlet reports that Meta has published weights for “Muse Spark Open,” positioning it as a less capable model than “Muse Wonder,” Meta’s more advanced frontier offering. Meta’s chief AI officer, Alexandr Wang, is cited as indicating that Meta generally does not open-source its most powerful models, even as it provides more transparency for lower-tier releases.
Another account frames Muse Wonder and a larger variant as Meta’s latest attempt to challenge leading rivals, noting the intended use for personal and professional AI agents. While both sets of models emphasize coding and reasoning, the coverage highlights a key difference in openness: lower-tier model weights are shared, while the most frontier capabilities remain gated.
