It’s raining frontier AI models. Soon after Anthropic set alarm bells ringing with its Claude Fable model at the summit of AI peformance, OpenAI answered with GPT-6 Astra. Not one to be left behind, Google released Gemini 4 Argon. The race is now getting hotter as Meta enters the fray with two new AI models in the Muse series: Wonder and Wonder Full. Cheesy naming aside, these Meta AI models mean business, or at least that’s what the benchmark figures suggest. More importantly, these AI models are already making their way to Muse users.
Beyond the benchmark figures, what sets these two models apart is a couple of user-facing upgrades: Active Memory and Active Focus. The former enables the model to learn and remember user preferences across different chats. The latter is a looping system that keeps trying different approaches when the first run fails, and keeps going until the task at hand is accomplished.
Meta says it has paid special attention to the safety aspect, which seems like a sane approach, given the recent string of unexpected (read: extremely risky and concerning) behavior by top-tier AI models developed by rivals like OpenAI, Google, and Anthropic. “That’s why we trained the Muse Wonder family to avoid sharing private information without permission; turn down requests to assist in performing actually harmful acts; and work to improve security and strengthen defenses against cyber attacks,” says the company.
Now, let’s talk performance. On the Artificial Intelligence Index, Muse Wonder Full only lags behind Claude Opus 5.5, while racing ahead of Claude Fable 5.1, GPT-6 Astra, and Gemini 4 Argon. On the Personal SuperIntelligence Index, Memory Bench, and Automation Bench, it sits right at the top. More importantly, in coding workflows, both the Muse Wonder models beat industry favorites — Claude Opus and GPT-6 — at their max performance levels.
That is no mean feat. Until a few months ago, Meta was not considered a true contender in the AI race, with Anthropic, OpenAI, and Google competing against each other at the vanguard of AI development. Meta’s comeback is quite remarkable, and it has already expanded Muse AI’s footprint dramatically, ranging from enterprise to consumer adoption through smart glasses and the upcoming Muse series hardware.