Meta introduced two new artificial intelligence models on Monday, Oct. 12, designed to enhance the capabilities of personal and professional AI agents. The company announced the release of Muse Wonder and Muse Wonder Full in a press release, marking a significant update to its existing AI infrastructure.
Muse Wonder became available to developers and builders through the Meta Model API and Muse Code. Muse Wonder Full was simultaneously released for Muse Code subscribers. According to the release, these new models will soon power additional features within Meta’s AI agent, Muse.
Active Memory and Focus Capabilities
The new models build upon the foundation of Muse Spark, which Meta introduced in April as part of its roadmap toward personal superintelligence. The company stated that Muse Wonder and Muse Wonder Full deliver state-of-the-art performance on agentic tasks, allowing the systems to interpret complex information and work through problems independently.
Both models are equipped with Active Memory, a feature that enables them to retain relevant details from past conversations. This allows the agents to maintain context over longer interactions without requiring users to repeat information. The models also feature Active Focus, which allows them to handle difficult tasks by attempting alternative approaches, recovering from errors, and asking for clarification when necessary.
Meta emphasized the practical applications of these features in expert domains. The company noted that the models are designed to assist with tasks such as navigating insurance claims, understanding lease agreements, and analyzing personal finances. The release highlighted that the models bring these capabilities to fields including law, finance, and health, while also providing frontier coding performance for developers building games or applications.
Safety and Strategic Positioning
Meta stated that the models were trained, evaluated, and released in accordance with the Meta Superintelligence Scaling Framework. The company specified that the models are trained to avoid sharing personal information without permission and to decline requests to assist in harmful acts. These safety measures are integrated into the core training process to ensure responsible deployment.
The launch of these models comes as Meta continues to expand the role of its Muse agent. Meta launched the original Muse platform on Sept. 8, describing it as a personal AI agent powered by Muse Spark. The company positioned Muse as a simple-to-use tool designed for a broad user base.
Industry analysts have noted the growing importance of Muse in Meta’s broader strategy. A Citi analysis reported by Reuters on Oct. 6 suggested that Muse is increasingly serving as a primary gateway to internet services. The analysis projected that the platform could generate revenues exceeding $27 billion by 2030, indicating Meta’s intent to place Muse at the center of its artificial intelligence efforts.
Developers interested in using Muse Wonder can access the model through the Meta Model API. Muse Code subscribers have immediate access to Muse Wonder Full. The integration of these models into the broader Muse ecosystem is expected to expand as development continues.
Source: PYMNTS
