Meta launches Muse Wonder and Muse Wonder Full models in the advanced models category.

Meta Launches Muse Wonder and Muse Wonder Full AI Models

Meta’s Superintelligence Lab has introduced two new artificial intelligence models, Muse Wonder and Muse Wonder Full. This marks the company’s first direct entry into the advanced model category, aiming for high-level performance in complex tasks.

Enhanced Capabilities and Benchmarking

The new models feature larger sizes and enhanced capabilities compared to the lightweight Muse Spark model. They deliver benchmark results comparable to Fable 5.1 and GPT-6 Astra, along with improved performance in programming and complex tasks.

Pricing for Meta Model API Access

• Muse Wonder: Costs four dollars per million input tokens and twenty dollars per million output tokens.
• Muse Wonder Full: This model will not be available through the API and will be exclusive to Muse Code subscribers.

Active Memory for Personal Assistant Enhancement

Both models will power the AI agent Muse. Meta aims to develop assistants capable of executing long-term tasks, with the ability to retain and retrieve information through what is termed active memory. The MemoryBench test demonstrated that Muse Wonder outperformed Fable, Opus, and Astra in continuous learning and memory building.

Strategic Shift in Meta’s Direction

The launch of these models represents the first release in the advanced model category since the company’s restructuring efforts in August, led by Alexander Wang. Through Muse Wonder, Meta seeks to compete directly in the advanced model market, particularly in programming, complex tasks, and long-term memory, with both models trained at the Prometheus data center in Ohio, USA.

The post Meta Launches Muse Wonder and Muse Wonder Full in the Advanced Model Category appeared first on Yemen TV.

To follow the news in Arabic