Meta Enters AI Frontier With New “Wonder” Models

Meta has entered the AI frontier with the release of its new "Wonder" models, Muse Wonder and Muse Wonder Full. These frontier AI models, announced by the Meta Superintelligence Lab, aim to compete with industry rivals OpenAI and Anthropic. The larger and more capable Wonder model outperforms Meta's earlier Muse Spark models and matches Anthropic's Claude Fable 5.1 and OpenAI's GPT-6 Astra in benchmarks.

According to Meta, the models represent a significant leap towards personal superintelligence. Wonder is available through the Meta Model API and Muse Code, while Wonder Full is limited to Muse Code subscribers. Both models offer a 1 million token context window and support text, images, video, and audio as input. They also boast "active memory" and "active focus" capabilities for retaining and processing user-shared information over time.

Independent scores from the Intelligence Index place Wonder Full slightly below Claude Opus 5.5 and GPT-6 Astra, but ahead of Anthropic's Claude Sonnet 5.5. Meta trained the models at its Prometheus data center in Ohio, marking the first frontier model launch since restructuring AI work following Alexandr Wang's leadership changes.

Developers can currently try Muse Wonder on the public API, while Wonder Full remains behind Muse Code's higher-tier plans.