Meta launches Muse Wonder AI models, claims frontier-tier performance

Meta announced on October 12 the release of Muse Wonder, a new family of large language models developed by its Superintelligence Labs. The models come in two versions: Muse Wonder, optimized for speed, and Muse Wonder Full, the complete version. According to a post by user dotey summarizing Meta's announcement, CEO Mark Zuckerberg said the model places Meta in the 'frontier' tier, alongside the strongest models from OpenAI, Anthropic, and Google. The post reports that developers can access Muse Wonder today through the Meta Model API or within Meta's AI coding tool Muse Code, while Muse Wonder Full is currently available only to Muse Code subscribers. The post notes a discrepancy: Meta's official blog says the consumer AI assistant Muse will integrate Muse Wonder 'soon' for subscribers, but Zuckerberg's X post claims subscribers can use Muse Wonder today. The model card specifies a 1 million token context window, multimodal input (text, images, video, audio) and text-only output. Meta's own benchmarks, as reported in the post, show Muse Wonder Full leading in agentic tasks, long-horizon coding, and legal tasks, but trailing Claude Opus 5.5 in several categories. The post also mentions that Meta has not disclosed pricing or open-weight plans, and that a full safety report is forthcoming.