Meta's New AI Chatbot Can Read Legal Docs and Write Code

Meta released Muse Wonder, a new AI model that Meta says is among the most capable in the world. The Muse Wonder models will be available through an API and through Muse Code, a tool for writing software with AI, and Meta says the best version will power features across its platforms, including its Muse personal agent — not as a free trial, but through those channels.

Meta says the largest version outperforms Anthropic's Claude Fable 5.1 and ranks just below Claude Opus 5.5 on the Artificial Analysis Intelligence Index, and that it beats OpenAI's GPT-6 Astra on some specialized benchmarks, including ones measuring work with legal documents and building software. Benchmarks aren't always a reliable measure of real-world performance, but the article calls this a significant step forward for Meta's models.

Muse Wonder also includes two new features, Active Memory and Active Focus, designed to keep the model locked into a stated goal and prevent it from becoming confused during complex research or large software projects. Meta says it trained the model to refuse dangerous outputs, and Zuckerberg said last month the company delayed the release to work on safety. Still, recent incidents involving rogue AI agents from Anthropic, OpenAI, and others show how easily powerful models can stray, and Meta did not disclose whether Muse Wonder agents have misbehaved in testing. It suggests Meta is once again competitive in the AI race.