Meta launches Muse Wonder AI models to power advanced personal and professional assistants

Meta rolled out its most advanced frontier artificial intelligence models on Monday, October 12, aiming to strengthen its position in the tech sector. The new software, initially known during testing as 'Watermelon', includes the Muse Wonder and Muse Wonder-Full editions. They were created by the Meta Superintelligence Team under the guidance of Scale AI co-founder Alexandr Wang. The launch follows the earlier release of Meta's Muse personal agent and a major company overhaul led by CEO Mark Zuckerberg.

Core Features

The systems stand out through two primary tools named 'Active Memory' and 'Active Focus'. Active Memory lets the system hold onto user details for weeks or months to better complete tasks across different chats. Active Focus helps the software stay on track with difficult jobs, test different solutions, and ask questions when stuck. Together, these tools let the models handle software coding, sort out complex financial or legal paperwork, and plan tasks like househunting. The software accepts text, image, video, and audio inputs, while producing text outputs, and features a context window of one million tokens.

Market Position

Independent benchmark tests place the Wonder-Full version at a score of 56 on the Intelligence Index, matching Anthropic's Claude Sonnet 5.5 and beating rival systems from Google and OpenAI. On specific software testing, Wonder-Full scored 57.5 on SWE-Marathon, outperforming competing tools, though it trailed on health queries and terminal tasks. The standard Muse Wonder model is priced for enterprise developers at four dollars per million input tokens and twenty dollars per million output tokens via the Meta Model API, while the Wonder-Full version remains restricted to Muse Code subscribers.

Context

Tech companies are racing to build AI agents that can handle multi-step work tasks rather than just answering simple chat queries. Meta is leveraging its massive user base across apps like WhatsApp and Instagram to distribute these tools directly to everyday consumers. However, outside developers and researchers will still need to verify Meta's performance claims, while privacy watchdogs continue to monitor how tech firms collect user data for training.