AI

Meta Launches Muse Wonder Models for Personal and Professional AI Agents

Monday, October 12, 2026Read Original

Details

  • Meta introduced Muse Wonder and Muse Wonder Full on October 12, 2026, describing them as its most capable models for personal and professional agentic tasks.
  • Muse Wonder is available through the Meta Model API and Muse Code, while Muse Wonder Full is restricted to Muse Code subscribers; both are expected to support future Muse experiences.
  • The models target tasks spanning insurance claims, leases, finances, law, health, coding, and app development, with Active Memory for cross-conversation continuity and Active Focus for persistence, recovery, and clarification.
  • Meta said the models were trained to protect private information, refuse harmful assistance, and help security teams identify software vulnerabilities under its Superintelligence Scaling Framework.
  • The Meta Model API lists Muse Wonder at $4 per million input tokens and $20 per million output tokens; Muse Wonder Full is subscription-only and is not offered through the API.[1][3][5]

Impact

The launch expands Meta’s push from conversational AI toward persistent agents that can execute multi-step work across files, messages, and applications. API access may broaden developer experimentation, while the subscription-only higher-capability model concentrates advanced coding and computer-use workflows inside Muse Code. Over the next 12–24 months, agent reliability, memory controls, privacy safeguards, and usage economics are likely to shape adoption more than benchmark claims.

Rift Dispatch