Meta launched Muse Wonder and Muse Wonder Full as its first frontier AI models, backed by internal benchmark tests showing strong performance in coding and agentic tasks.
Developers can access the models via the Meta Model API and Muse Code.
Meta has published internal test results comparing its new models to those from OpenAI and Anthropic. The data suggests that the more complete version can compete with or even surpass the competition in terms of personal intelligence, scheduling, automation, and memory. However, these are measurements made by the company itself, and their real-world performance remains to be seen.
The basic version is already available to those building apps using Meta's tools, while the more powerful model is available to Muse Code subscribers. Soon, this technology will also be utilized in Meta's digital tool, Muse AI.
The change of course and the company's strategy
The development of the new models comes after a major shift in the company's strategy that began 14 months ago. During the same period, Meta spun off Scale AI and appointed its co-founder as head of artificial intelligence.
Meta CEO Mark Zuckerberg said that creating these systems required huge resources, investment and intensive scientific work. At the same time, he chose not to join the group of technology company leaders calling for a halt to the development of artificial intelligence, emphasizing that his company places more emphasis on personal empowerment of users compared to others that focus on professional tasks and automation.
Top comments (0)